Thoth-Mem
Thoth-Mem
Permanenter Speicher für KI-Codierungsagenten
Gib Codierungsagenten dauerhaften Projektspeicher über Sitzungen, Kompaktierungen und Kontextzurücksetzungen hinweg.
Thoth-Mem ist ein lokaler MCP-Server auf Basis von SQLite und FTS5. Er bewahrt nützliche Entscheidungen, Fehlerbehebungen, Konventionen und Sitzungskontinuität auf und ruft dann nur die Beweise ab, die ein Agent benötigt. Dieselbe Installation bietet außerdem eine CLI, eine optionale HTTP-API und native Lebenszyklus-Integrationen für unterstützte Codierungs-Harnesses.
Der globale Geltungsbereich verwaltet die Harness-Konfiguration des aktuellen Benutzers; der Projekt-Geltungsbereich ist explizit und auf das ausgewählte Projekt und seinen Belegbaum beschränkt. Engram, thoth-agents oder eine andere Speicherintegration können sich überschneiden; betrachten Sie dies nur als Warnung: thoth-mem bearbeitet, deaktiviert oder entfernt keine externen Repositorys und schreibt nicht in sie.
Schnellstart
Erfordert Node.js 18 oder neuer. Das native Setup ist optional: Eine manuelle MCP-Verbindung benötigt nur den Befehl mcp.
Das veröffentlichte Paket ausführen
Starten Sie den neuesten veröffentlichten MCP-Server, ohne einen globalen Befehl zu installieren:
npx -y thoth-mem@latest mcpDies startet den MCP-Server und seine lokale HTTP-Brücke. Fügen Sie --no-http hinzu, wenn nur der MCP-Transport gewünscht ist. Neue Client-Konfigurationen sollten den expliziten Unterbefehl mcp verwenden.
Native Integrationen rufen nach dem Setup den persistenten Befehl thoth-mem auf. Installieren oder aktualisieren Sie diesen Befehl daher global, bevor Sie eine Harness konfigurieren. Verwenden Sie npx, um die Setup-Implementierung aus dem neuesten veröffentlichten Paket auszuführen, prüfen Sie ihren Plan ohne Schreibvorgänge und wenden Sie ihn dann an:
npx -y thoth-mem@latest setup codex --scope global --plan --json
npx -y thoth-mem@latest setup codex --scope global --jsonErsetzen Sie codex durch opencode oder claude für eine andere unterstützte Harness und starten Sie diese Harness dann neu. Die alleinige Ausführung von setup installiert oder aktualisiert das npm-Paket nicht.
Dieses Repository installieren
Verwenden Sie den Repository-Ablauf, um Commits zu testen, die noch nicht veröffentlicht wurden:
pnpm install
pnpm run build
pnpm add -g .
thoth-mem version
thoth-mem setup codex --scope global --plan --json
thoth-mem setup codex --scope global --jsonthoth-mem@latest enthält nur die neueste veröffentlichte Version. Bauen Sie neu und führen Sie pnpm add -g . erneut aus, nachdem Sie neuere unveröffentlichte Commits gezogen haben.
Eine bestehende Installation aktualisieren
Aktualisieren Sie zuerst das Paket. Wenn eine native Integration installiert ist, führen Sie deren Setup erneut aus, damit kopierte Assets, Skills, Hooks und verwaltete Deklarationen mit der neuen Paketversion konvergieren:
pnpm add -g thoth-mem@latest
npx -y thoth-mem@latest setup codex --scope global --plan --json
npx -y thoth-mem@latest setup codex --scope global --jsonStarten Sie dann die Harness oder den MCP-Prozess neu. Manuelle MCP-Benutzer benötigen kein setup; ein Neustart von npx -y thoth-mem@latest mcp genügt.
Das Setup bewahrt die Speicherdatenbank und die benutzereigene Konfiguration. Beim Start können fehlende Konfigurationsfelder nachgefüllt werden, aber explizite Werte wie ein LM-Studio-Modell bleiben ausgewählt. Das Konfigurationsformat bleibt "version": 1. Aktualisieren Sie bei einer veröffentlichten Installation eine ältere $schema-URL manuell auf diese Release-Version für aktuelle Editor-Validierung und Autovervollständigung. Ein unveröffentlichter Checkout muss config.schema.json dieses Repositorys für passende Validierung verwenden, da unpkg die Änderung vor der Veröffentlichung nicht bereitstellen kann. Die Schema-URL steuert keine Laufzeit-Migration.
Das Ändern eines Embedding-Modells ist ein Konfigurationsvorgang, kein Setup-Vorgang. Bearbeiten Sie embedding.provider, model, baseUrl und native dimensions nach Bedarf; profile: "auto" löst unterstützte Modellfamilien auf. Starten Sie thoth-mem neu und lassen Sie die geänderte Embedding-Abstammung den idempotenten Neuaufbau des semantischen Index in die Warteschlange stellen.
Related MCP server: LumenCore
Die Speicherschleife
Ein nützlicher Agenten-Workflow ist klein und wiederholbar:
Die dauerhafte Lektion speichern. Verwenden Sie
mem_savefür eine Entscheidung, Grundursache, Konvention oder andere nicht offensichtliche Tatsache, die den aktuellen Kontext überleben soll.Eng abrufen. Beginnen Sie mit
mem_recall(mode="compact"), erweitern Sie starke Kandidaten mitmode="context"und rufen Sie einen vollständigen ausgewählten Datensatz mitmem_getab.Mit Identität fortfahren. Behalten Sie dieselbe stabile
session_idund dasselbeproject; verwenden Siemem_contextfür aktuelle Kontinuität undmem_sessionfür Root-eigene Lebenszyklus-Ereignisse.
Beispiel-Beobachtung:
{
"kind": "observation",
"title": "Retry SQLite writes in a new transaction",
"type": "bugfix",
"project": "my-project",
"topic_key": "sqlite/busy-retry",
"content": "**What**: Roll back after SQLITE_BUSY and retry in a new transaction.\n**Why**: Retrying inside the failed transaction repeats the failure.\n**Where**: write transaction helper.\n**Learned**: Use bounded backoff before opening the new transaction."
}Entfernen Sie Inhalte innerhalb von <private>...</private> vor der Persistierung. Speichern Sie keine Anmeldeinformationen, vollständigen Transkripte, generierten Agenten-Prompts als Benutzerabsicht oder Rohprotokolle ohne eine wiederverwendbare Lektion.
Sechs MCP-Tools
Tool | Verwendung |
| Eine Beobachtung, echte Benutzer-Prompt, Root-eigene Zusammenfassung oder passives Lernen persistieren. |
| Begrenzten fusionierten Abruf ausführen; kompakte Ergebnisse verwenden, bevor der Kontext erweitert wird. |
| Aktuelle Sitzungen, Prompts, Beobachtungen und optionale abgerufene Kontinuität lesen. |
| Eine Beobachtung oder einen Prompt nach ID abrufen, mit begrenzter Paginierung oder Zeitachsenkontext. |
| Projekte, Themen, Graph-Ansichten und betriebliche Gesundheit navigieren. |
| Eine Root-eigene Speichersitzung starten, einen Prüfpunkt setzen oder zusammenfassen. |
Setup-, Synchronisierungs-, Migrations-, Neuaufbau- und Wartungsbefehle sind CLI/HTTP-Verwaltung, keine zusätzlichen MCP-Tools.
Graph-Communities untersuchen
Communities sind begrenzte Zusammenfassungen, die aus dem Wissensgraph eines Projekts abgeleitet werden. Ein Operator erstellt oder aktualisiert die festgeschriebenen Zusammenfassungen über die CLI:
thoth-mem rebuild-communities --project my-projectEin Agent erhält sie dann über mem_project:
{
"action": "graph",
"project": "my-project",
"navigation": "community",
"limit": 5,
"max_chars": 2000
}Die Antwort meldet Community-Zustand und Aktualität, dann Einträge wie community=<id>, Graph-Abdeckung, Konfidenz, Degradationszustand, eine begrenzte Zusammenfassung und sources=obs:<id>. Die Community-Untersuchung erfordert ein Projekt, aber keinen Fokus-Knoten oder Beobachtungs-ID. Wenn keine festgeschriebenen Zusammenfassungen existieren, wird dies mitgeteilt, anstatt eine globale Antwort zu synthetisieren.
Um Beweise hinter einer Community zu untersuchen, nehmen Sie ein obs:<id> aus deren sources-Feld und rufen Sie mem_get(kind="observation", id=<id>) auf. Beobachtungs-IDs erscheinen auch in Abrufergebnissen. Für eine begrenzte Graph-Nachbarschaft verwenden Sie eine als focus_node_id="obs:<id>" mit navigation="neighborhood" erneut.
Native Harness-Integrationen
Das native Setup installiert die gepackte MCP-Deklaration, den Speicher-Skill und Lebenszyklus-Hooks, wo die Harness diese unterstützt. Prüfen Sie zuerst den Plan ohne Schreibvorgänge und führen Sie ihn dann ohne --plan erneut aus, um ihn anzuwenden:
Harness | Plan | Anwenden |
OpenCode |
|
|
Codex |
|
|
Claude Code |
|
|
Der Standard-Setup-Befehl für OpenCode ist thoth-mem setup opencode; fügen Sie thoth-mem setup opencode --scope project --project /path/to/project --force hinzu, wenn Sie explizit ein Projekt ansprechen, oder verwenden Sie thoth-mem setup codex --rollback /path/to/receipt.json für ein belegbezogenes Rollback.
Setup-Status und Prozess-Exit-Codes sind stabil:
Status | Exit-Code |
|
|
|
|
|
|
|
|
Projektlokales Setup ist explizit:
thoth-mem setup opencode --scope project --project /path/to/project --plan --jsonPrüfen Sie erkannte Konflikte vor dem Anwenden. Verwenden Sie --force nur für konfliktbehaftete Speicherorte, deren thoth-mem-Eigentum bereits nachgewiesen ist. Codex 0.144.x, 0.146.x und 0.147.x gehören zum getesteten Kompatibilitätssatz und erfordern kein --force. Für andere Codex-Versionen darf --force nur die Testversions-Sperre überschreiben, wenn der ausgewählte Geltungsbereich weiterhin vollständige, unabhängig überprüfbare Plugin-Manager-Fähigkeiten bereitstellt; das Setup gibt eine Warnung aus, wenn es diese Überschreibung verwendet. Es umgeht keine Zustandsprüfung, Eigentums-, Eindämmungs-, Abgleich- oder Bereinigungsschutzmaßnahmen und gewährt keine Autorität über unzusammenhängende Konfiguration.
Claude Code unterstützt auch seinen nativen Marketplace-Ablauf:
claude plugin marketplace add EremesNG/thoth-mem
claude plugin install thoth-memDie native Integration ist optional. Bestehende Speicher und der MCP-Server mit sechs Tools funktionieren weiterhin mit einer manuellen Verbindung.
Manueller MCP-Fallback
Native Hooks sind optional. Behalten Sie eine einfache MCP-Verbindung mit sechs Tools bei, wenn Sie kein verwaltetes Setup oder natives Plugin wünschen; bestehende Speicher bleiben verfügbar.
Übergang zur nativen Harness-Integration
Das native Setup ist optional: Prüfen Sie den Plan ohne Schreibvorgänge, überprüfen Sie Konflikte und wenden Sie dann den passenden Harness-Befehl an. Öffnen Sie für Codex /plugins, installieren Sie thoth-mem von EremesNG/thoth-mem und verifizieren Sie den Marketplace- und Plugin-Zustand. Die externe Codex-Registrierung ist nicht atomar umkehrbar. Bestätigen Sie daher den externen Zustand, bevor Sie das lokale Setup erneut versuchen oder zurückrollen.
Vertrag für verwaltetes Setup: Der Planmodus führt keine Schreibvorgänge aus und mutiert nur an thoth-mem-verwalteten Speicherorten. Backups werden vor der ersten Mutation erstellt; OpenCode akzeptiert opencode.json oder opencode.jsonc. Jeder mutierende Versuch schreibt einen HMAC-geschützten Beleg mit Status in_progress vor Änderungen:
globale Belege:
<thoth-data-dir>/setup/receipts/<receipt-id>/receipt.jsonProjekt-Belege:
<project>/.thoth/setup/receipts/<receipt-id>/receipt.json
Fehlende oder manipulierte Belege schlagen sicher fehl. Ein verifiziertes Rollback bewahrt unzusammenhängende Einstellungen, während Abweichungen oder nicht verfügbare Fähigkeiten requires_user_action zurückgeben. Wiederholtes Setup und wiederholtes abgeschlossenes Rollback sind No-Ops, wenn der verifizierte Zustand bereits übereinstimmt.
Gemini CLI: manuelles MCP
Gemini CLI ist ein manueller MCP-Client-Pfad, keine verwaltete native thoth-mem-Integration. Fügen Sie diesen Eintrag zu ~/.gemini/settings.json hinzu:
{
"mcpServers": {
"thoth": {
"command": "npx",
"args": ["-y", "thoth-mem@latest", "mcp"]
}
}
}Abruf- und Graph-Qualität bewerten
Das Repository enthält deterministische Bewertungsbefehle:
pnpm run eval:retrieval
pnpm run eval:kg
pnpm run eval:embedding-models -- --helpeval:retrieval setzt Signal-Beobachtungen plus Distraktoren und misst, ob der erwartete Speicher nahe der Spitze rangiert. Lesen Sie seinen Bericht als Sammlung von Signalen:
Recall und Rang zeigen, ob die richtigen Beweise gefunden wurden und wie früh.
Rauschen und Fallmischung zeigen Robustheit über direkte, umformulierte und repository-abgeleitete Beispiele.
Kompression zeigt, wie viele Beweise vor der Kontextzustellung entfernt wurden; es ist ein Effizienzsignal, kein Beweis, dass der verbleibende Text korrekt ist.
Spur- und Fallback-Beweise zeigen lexikalische, semantische Roh-/HyDE- und KG-Beteiligung, einschließlich ausstehendem oder degradiertem semantischem Verhalten.
Abstammung und Herkunft zeigen, ob zurückgegebene Beweise ihrer Quelle zuordenbar bleiben.
eval:kg misst erwarteten Subjekt-Relation-Objekt-Recall, verbotene Tripel-Leckage, deterministisches Extraktionsverhalten und validierte optionale LLM-Anreicherung. Fehlende erwartete Fakten weisen auf Abdeckungslücken hin; verbotene Treffer weisen auf unsichere Graph-Erfindung hin.
Diese Bewertungen sind deterministische Entwicklungs-Gates über kuratierte und synthetische Fixtures. Sie sagen nicht jeden Produktionskorpus voraus, ersetzen keine menschliche Überprüfung, beweisen keine native Harness-Integration oder rechtfertigen für sich genommen die Aktivierung optionaler Community-Lesepfade. Vergleichen Sie die einzelnen Fälle und Fehlermeldungen, anstatt eine aggregierte Zahl als universelle Qualität zu behandeln.
Embedding-Profile und Modellvergleich
Semantische Embedding-Eingaben werden von einem versionierten Modellprofil formatiert. auto erkennt Nomic-, EmbeddingGemma- und Qwen3-Embedding-Modellfamilien-Aliase; unbekannte Modelle verwenden raw und erhalten keine abgeleitete asymmetrische Formatierung. Die öffentliche Konfiguration hat absichtlich kein globales task-Feld: Abrufabsicht und Abfrage-/Dokumentrolle werden intern für jede Eingabe zugewiesen, einschließlich HyDE-Antworten in Dokumentrolle.
{
"embedding": {
"provider": "lmstudio",
"model": "text-embedding-embeddinggemma-300m",
"baseUrl": "http://127.0.0.1:1234",
"dimensions": 768,
"profile": "auto",
"normalize": true
}
}Unterstützte Profilwerte sind auto, nomic, embeddinggemma, qwen3 und raw. THOTH_EMBEDDING_PROFILE und THOTH_EMBEDDING_NORMALIZE überschreiben gespeicherte Werte. Die aufgelöste Profilversion und das Normalisierungsflag sind Teil der Abstammung des semantischen Index. Eine Änderung dieser Werte markiert daher frühere Vektoren als veraltet und nutzt die bestehende idempotente Neuaufbau-Warteschlange.
Lokale Transformers.js-Inferenz kann sich für ein bestimmtes ONNX-Ausführungsgerät entscheiden:
{
"embedding": {
"provider": "transformers_local",
"model": "onnx-community/embeddinggemma-300m-ONNX",
"device": "dml",
"dimensions": 768,
"profile": "auto",
"normalize": true
}
}Unterstützte Gerätewerte sind auto, cpu, dml, cuda und coreml; cpu ist die Standardeinstellung. THOTH_EMBEDDING_DEVICE überschreibt den gespeicherten Wert von embedding.device. Mit der von Transformers.js verwendeten vorgefertigten Node-ONNX-Runtime zielt dml auf DirectML unter Windows, cuda auf unterstützte Linux-x64-CUDA-Installationen und coreml auf macOS. Ein explizit nicht verfügbares Gerät führt zu einem Fehler bei der Modellinitialisierung, anstatt stillschweigend auf CPU umzuschalten. auto delegiert die plattformspezifische Provider-Reihenfolge und das Fallback an Transformers.js, sodass sich das effektive Backend je nach Host oder Abhängigkeitsversion ändern kann.
Die Geräteauswahl betrifft nur transformers_local; Remote-Anfragen an Ollama und LM Studio ignorieren sie. GPU-Backends können einen deutlich langsameren Kaltstart haben und sind daher am nützlichsten für persistente MCP-Prozesse oder größere Embedding-Stapel. Das Gerät ist bewusst von der Abstammung des semantischen Index ausgeschlossen: Eine Änderung nur von embedding.device markiert vorhandene Vektoren nicht als veraltet und stellt keinen Neuaufbau in die Warteschlange.
Beispiele für Provider-Modelle:
Profil | LM-Studio-Modell-ID | Transformers.js-Modell-ID | Native Dimensionen |
Nomic | verwenden Sie die exakte ID aus |
| 768 |
EmbeddingGemma |
|
| 768 |
Qwen3-Embedding-0.6B |
|
| 1024 |
Die lokale Ausführung von EmbeddingGemma verwendet sentence_embedding. Die lokale Qwen-Ausführung wendet die Retrieval-Anweisung nur auf Abfragen an und verwendet das zuletzt beachtete Hidden-State-Token für das Pooling. Alle Provider lehnen unvollständige, nicht endliche, Null- oder dimensional inkonsistente Stapel ab. LM-Studio-Antwortindizes werden validiert, und gültige, nicht in Reihenfolge befindliche Zeilen werden in die Eingabereihenfolge zurückversetzt; fehlende, doppelte oder ungültige Indizes werden abgelehnt. Während des Abrufs stufen diese Fehler die semantische Suche explizit herab, während lexikalische und KG-Suche fortgesetzt werden.
Führen Sie das Qualitätsgate für drei Modelle mit expliziten Modell-IDs und einem dauerhaften Ausgabepfad aus:
pnpm run eval:embedding-models -- --provider lmstudio --base-url http://127.0.0.1:1234 --nomic-model <nomic-id> --embeddinggemma-model <gemma-id> --qwen3-model <qwen-id> --output <result.json>Das Gate erfordert, dass alle drei Ausführungen abgeschlossen werden und mindestens ein Kandidat die Recall@1/Recall@5/MRR-Schwellenwerte erfüllt, ohne eine Nomic-Metrik zu verschlechtern. Nomic ist der relative Vergleichsmaßstab, kein Kandidat, der den absoluten Schwellenwerten unterliegt. Wenn beide Kandidaten qualifiziert sind, wählen ein expliziter Qualitätswert und eine stabile Tie-Break-Reihenfolge den Gewinner. Ein fehlendes Modell, ein ungültiger Vektor, kein geeigneter Kandidat oder ein Fehler beim Schreiben des Berichts führt zu einem Exit-Code ungleich Null und behält die aktuelle Standardeinstellung bei.
Der aufgezeichnete LM-Studio-Lauf vom 2026-08-08 wählte EmbeddingGemma als ausgelieferten lokalen Standard. EmbeddingGemma und Qwen3 wurden beide mit Recall@1 1.00, Recall@5 1.00 und MRR 1.00 abgeschlossen, gegenüber Nomic mit 0.50, 1.00 und 0.7167; beide Kandidaten waren geeignet, und EmbeddingGemma gewann den exakten Qualitätsgleichstand durch die stabile lexikalische Profil-ID-Regel. Die mittlere Latenz im gespeicherten Entscheidungslauf betrug 190.5 ms für Nomic, 195 ms für EmbeddingGemma und 320.5 ms für Qwen3.
Die Dateigrößen von Qwen3 hängen vom Laufzeit-Artefakt ab:
Qwen3-Artefakt | Quantisierung | Bytes | MiB |
Original Transformers | BF16 | 1,191,586,416 | 1,136.39 |
Transformers.js | Q8 | 613,527,631 | 585.11 |
LM Studio | Q8_0 | 639,150,592 | 609.54 |
Das Qwen3-Q8-Modell ist in Transformers.js 304.069.133 Bytes größer als EmbeddingGemma Q8 und in LM Studio 305.559.648 Bytes größer. Es verwendet außerdem native 1024-dimensionale Vektoren anstelle der 768 Dimensionen von EmbeddingGemma. Der Runner installiert oder erkennt keine Provider-Modelle im Namen des Betreibers.
Skalieren Sie das Retrieval-Rauschen, wenn Sie einen anspruchsvolleren lokalen Lauf wünschen:
$env:THOTH_RETRIEVAL_EVAL_NOISE='250'
pnpm run eval:retrievalErweiterte Operationen
Führen Sie
thoth-mem helpaus, um die vollständige Liste der CLI-Befehle und Optionen zu erhalten.Öffnen Sie das lokale Dashboard unter
http://localhost:7438/und die OpenAPI-Dokumentation unterhttp://localhost:7438/docs.Verwenden Sie
thoth-mem sync --dir=.thoth-syncundthoth-mem sync-import --dir=.thoth-syncfür Git-freundliche Portabilität.repair-sync-journal (--project <name> | --all) --applyzeigt eine Vorschau an und bindet seinen Reparaturstapel intern. Das optionale--expected-fingerprintbleibt verfügbar, wenn ein externer Workflow bereits eine Vorschau-Bindung hat.prune-operation-traces (--project <name> | --all) --applybindet ebenfalls einen Aufbewahrungsstapel intern. Fügen Sie--until-completehinzu, um den anfänglich begrenzten Rückstand mit einem festen effektiven Zeitpunkt und neuen späteren Fingerabdrücken zu verarbeiten. Eine extern bereitgestellte Bindung muss sowohl--expected-fingerprintals auch--effective-nowenthalten.compact-database [--data-dir <path>]führt eine schreibgeschützte Vorschau durch. Fügen Sie--applyerst hinzu, nachdem Sie die Schätzungen zu wiedergewinnbarem Speicherplatz und Kapazität geprüft haben.--applykann das Doppelte der größeren der physischen und logischen Datenbankgrößen erfordern, kann durch andere SQLite-Clients blockiert werden und meldet Erfolg erst nach Integritäts-, Fremdschlüssel-, Schema-, Dauerzähl- und WAL-Prüfungen. Es verwendet den von SQLite verwalteten Checkpoint undVACUUM; es verspricht kein Rollback nach einer festgeschriebenen Komprimierung.Die Komprimierung erfolgt niemals automatisch. Die Ausführung gegen Live-Daten erfordert eine separate Autorisierung des Betreibers; Repository-Tests verwenden nur Wegwerf-Datenbanken.
Prüfen Sie
config.schema.jsonfür gespeicherte Konfiguration und umgebungsgestützte Einstellungen.Daten liegen standardmäßig in
~/.thoth/thoth.db; überschreiben Sie das Datenverzeichnis mitTHOTH_DATA_DIRoder--data-dir.
Die semantische Indizierung ist nicht blockierend. Wenn Embeddings oder sqlite-vec nicht verfügbar sind, bleibt der Abruf über unterstützte lexikalische und grafische Evidenz nutzbar und meldet den herabgestuften Pfad, anstatt stillschweigend semantischen Erfolg zu behaupten.
Entwicklung
pnpm install
pnpm run integration:verify
pnpm run build
pnpm testLizenz
Available Tools
6 toolsmem_contextC
Build bounded handoff-first continuity. project_key required; root_session_key and harness together for session context, neither for project context.
| Name | Required | Description | Default |
|---|---|---|---|
| harness | No | Verified native harness. | |
| project_key | Yes | Verified project_key, copied verbatim; never derived. | |
| budget_chars | No | ||
| correlation_id | No | ||
| finalize_answer | No | ||
| root_session_key | No | Verified root session key. |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description bears the full burden. It implies a build/continuity operation but does not state whether it writes data, what side effects or permissions exist, whether it is idempotent, or what the result contains.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two compact sentences with no filler, and the required-parameter guidance is front-loaded. The only weakness is that the opening sentence is so terse and jargon-heavy that it obscures meaning.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a six-parameter tool with no annotations and no output schema, the description is too thin. It explains the session/project context split but omits behavioral details, output expectations, and several parameter meanings.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The description adds important conditional semantics for project_key, root_session_key, and harness, including the session-vs-project mode distinction. However, with only 50% schema description coverage, it leaves budget_chars, correlation_id, and finalize_answer unexplained.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description says it builds 'bounded handoff-first continuity,' which gestures at context creation but is vague jargon rather than a clear verb+resource statement. It also does not distinguish this tool from siblings such as mem_recall, mem_session, or mem_project.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It specifies project_key is required and explains that root_session_key plus harness together select session context while neither selects project context. That is useful conditional guidance, but it gives no explicit when-to-use-this-tool-vs-alternatives guidance or exclusions relative to sibling tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
mem_getB
Expand one selected record from a prior tool result.
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | Memory, summary, observation, or evidence id. | |
| history | No | Include predecessors; evidence has none. | |
| correlation_id | No |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It does not state that this is a read-only retrieval, does not explain what expansion returns, does not mention the optional history behavior, and gives no permission or side-effect context. The word 'expand' hints at non-destructive retrieval but that is the only signal.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single focused sentence with no wasted words, front-loading the core action. It is perhaps too terse for the gaps it leaves, but structurally it is efficient and readable.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With no annotations and no output schema, the description is the only source of behavioral context, and it says nothing about return shape, history semantics, or the undocumented correlation_id. For a retrieval tool that may return variable content, this leaves the agent without enough to call it confidently.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 67%: the schema documents 'id' and 'history' with useful detail, while 'correlation_id' is undocumented. The description's only contribution is that 'id' refers to a record from a prior tool result, which adds marginal context beyond the schema. Baseline 3 fits given the partial coverage.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The verb 'expand' plus the object 'one selected record from a prior tool result' makes the purpose reasonably clear: fetch the full detail of a single record referenced by a previous result. It does not differentiate itself from the sibling mem_recall, so an agent cannot tell from the text alone which retrieval tool to pick.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The phrase 'from a prior tool result' implies the context of use — after another tool has returned a record stub — which gives some usage guidance. However, it never states when to use this vs. mem_recall or mem_context, and offers no exclusions or prerequisites.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
mem_projectC
Read-only project views. list: no fields required; returns at most 256 exact aliases/project with aliasCount/aliasesTruncated. timeline: project_key required; accepts only since, until, cursor, limit, budget_chars. briefing/summaries/observations: project_key required; optional session filter selects summaries/observations, not briefing's project-wide memories. history: id required from a prior result.
| Name | Required | Description | Default |
|---|---|---|---|
| id | No | ||
| limit | No | Timeline/observations: 1-100 items. | |
| since | No | Inclusive lower ISO-8601 instant; <= until. | |
| state | No | ||
| until | No | Inclusive upper ISO-8601 instant. | |
| action | Yes | ||
| cursor | No | Unchanged nextCursor; preserve project_key, since, until. | |
| harness | No | ||
| temporal | No | Summaries/observations: current=latest summaries/correction leaves; history=predecessors. | |
| project_key | No | Verified project_key, copied verbatim; never derived. | |
| budget_chars | No | Aggregate chars; timeline: positive integer clamped to 1024-20000. | |
| root_session_key | No | Session filter; pair with harness. |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are present, so the description carries the full burden. It does disclose read-only behavior and several useful constraints: list returns at most 256 aliases with aliasCount/aliasesTruncated, timeline accepts only a specific set of filters, and a session filter affects summaries/observations but not briefing. However, permissions, return shapes for most actions, and error behavior are not described.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is short and front-loads the read-only nature. It is dense and semicolon-delimited, making it harder to scan than a bulleted structure, but every clause conveys actionable information with little waste.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The tool has six actions and twelve parameters with no output schema, yet the description explains return values only for list. It does not explain when to use briefing versus summaries or observations, what those actions return, or how temporal/state/harness parameters affect behavior. Significant gaps remain for an agent to invoke the tool correctly across all actions.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 67%, so the schema already documents many parameters. The description adds action-specific constraints not obvious from the schema: timeline accepts only since/until/cursor/limit/budget_chars, list requires no fields and returns a capped alias set, history id must come from a prior result, and the session filter behavior varies by action. It leaves state, harness, temporal, and root_session_key semantics largely unaddressed beyond the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Starts with 'Read-only project views,' which is a vague resource description, then lists action names and parameter requirements. The action list gives some clarity about what the tool can do, but the overall purpose ('project views') is not concretely defined and the tool is not distinguished from siblings like mem_context or mem_recall.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Provides per-action parameter requirements (e.g., timeline requires project_key and accepts only certain filters), but gives no guidance on when to use this tool versus alternatives or which action to choose for a given intent. There are no exclusions or alternative-tool references.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
mem_recallB
Call at the start of non-trivial repository work before exploring, with task key terms, unprompted. Search promoted project memory before expanding selected records.
| Name | Required | Description | Default |
|---|---|---|---|
| mode | No | compact (default): snippets; context: selected content. | |
| limit | No | ||
| query | Yes | Non-empty task search terms. | |
| temporal | No | current (default) or historical guidance. | |
| project_key | Yes | Verified project_key, copied verbatim; never derived. | |
| budget_chars | No | ||
| correlation_id | No | ||
| finalize_answer | No |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden; it does disclose a two-stage retrieval workflow (search snippets, then expand selected records) and that results are pre-promoted memory. However, it says nothing about what finalize_answer, correlation_id, or budget_chars do, what the response looks like, or whether the operation has any side effects. Useful workflow context, but not enough for an 8-parameter tool without annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two short sentences with the imperative and timing front-loaded; nothing is filler. The phrase 'with task key terms, unprompted' is slightly awkward but still parses as useful guidance.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with 8 parameters, no annotations, and no output schema, the description covers purpose and timing but omits the semantics of most parameters and the shape of results. An agent would need to open the schema and still lack clarity on finalize_answer and correlation_id.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
At 50% schema coverage, the description must compensate, and it only hints at 'query' ('with task key terms') and implicitly at mode via 'expanding selected records'. The remaining parameters (project_key, limit, temporal, budget_chars, correlation_id, finalize_answer) get no explanation in the description beyond the schema's own text.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description names a specific action (search promoted project memory) and adds scope ('promoted', 'project memory') that is more than a restatement of the tool name. It doesn't explicitly contrast with siblings like mem_context or mem_get, leaving the agent to infer the boundary.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It gives a clear trigger ('at the start of non-trivial repository work, before exploring') and says to call it unprompted, plus a follow-up step ('before expanding selected records'). It stops short of naming which sibling to use instead when the task isn't repository work.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
mem_saveA
Save unprompted as soon as decisions, root causes, failed approaches, or non-obvious conventions are verified. Save durable discoveries and continuation handoffs. Direct promoted memory (not handoff): memory.content uses labeled Result, Rationale, Scope, Caveat / safe action lines; omit inapplicable Scope/Caveat and never invent details. Handoffs: Objective, Completed, First pending action, Blockers, Key files/checks; require a stable workstream topic_key; close a finished handoff by saving its outcome under the same topic_key. Send exactly one branch: evidence, observation, observation_review, or observation_promotion. Supply root_session_key and harness together or omit both.
| Name | Required | Description | Default |
|---|---|---|---|
| memory | No | Direct promoted memory; requires evidence without metadata. | |
| harness | No | Verified native harness. | |
| evidence | No | Compact factual evidence; metadata requires event_key + session pair and forbids memory. | |
| event_key | No | Stable idempotency key. | |
| observation | No | Supported candidate, not promoted; requires event_key. Session scope needs session pair + coverage; project scope forbids coverage. | |
| project_key | Yes | Copy opaque project_key verbatim from verified native identity; never derive from names, paths, remotes, branches, worktrees, host IDs, listings, or recalled content. | |
| project_name | Yes | Display/creation name, not identity; prefer persisted lifecycle/project name. | |
| root_session_key | No | Verified root session key; blank is absent. | |
| observation_review | No | Review supported observation; requires event_key + session pair. | |
| observation_promotion | No | Promote accepted observation without new prose; requires event_key + session pair. |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the burden well: it discloses branch mutual exclusivity, the root_session_key/harness all-or-nothing pairing constraint, and the requirement of a stable topic_key to close a handoff. It does not say what happens on supersede/dedup or what the call returns, leaving those gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with the trigger condition, and every clause carries a rule rather than filler. Density is high and semicolon-chained, with terse terms like 'safe action lines' that take re-reading, but nothing is expendable.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 10-parameter, deeply nested write tool with no annotations and no output schema, the description covers branch selection and the session-pair constraint but omits return behavior, supersede semantics, and how the four branches relate to each other downstream. Adequate to make a correct call, not complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so the baseline is 3, but the description adds meaning the schema cannot express: the labeled Result/Rationale/Scope/Caveat shape of memory.content and the Objective/Completed/First pending action/Blockers structure of a handoff. It leaves supersedes_id, outcome, and the memory-vs-evidence relationship to the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The imperative verb 'Save' plus enumerated resource types (decisions, root causes, failed approaches, conventions, handoffs) makes clear this tool persists memory. It does not explicitly contrast itself with the read-oriented siblings (mem_get, mem_recall, mem_context), but the write framing is unmistakable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Strong trigger guidance: 'Save unprompted as soon as... verified', plus durable discoveries and continuation handoffs. It also gives branch-selection rules ('Send exactly one branch'). It never names a sibling tool as an alternative or states a when-not-to-save condition, which keeps it off a 5.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
mem_sessionB
Record verified root lifecycle events, never ordinary saves. guide_post_compact content carries host compaction text.
| Name | Required | Description | Default |
|---|---|---|---|
| content | No | ||
| harness | Yes | Verified native harness. | |
| summary | No | Session summary. | |
| event_key | Yes | Stable lifecycle key; retries must resend identical content/summary. | |
| operation | Yes | Root lifecycle operation. | |
| project_key | Yes | Verified project_key, copied verbatim; never derived. | |
| project_name | Yes | Display/creation name, not identity; prefer persisted lifecycle/project name. | |
| root_session_key | Yes | Stable verified root session key. |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden, but it only says these are lifecycle events and that guide_post_compact content carries host compaction text. It does not disclose mutation semantics, idempotency expectations, permission/auth requirements, or what happens for each lifecycle operation beyond what the schema implies.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two short sentences, front-loads the core action, and contains no filler. It is efficient, though its brevity contributes to gaps elsewhere rather than being a structural flaw itself.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
This is a complex tool with eight parameters, nested summary objects, six lifecycle operations, and no annotations or output schema. The description does not explain the lifecycle operations, what is recorded or changed, retry behavior, or the relationship between required fields, leaving too much for the agent to infer.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is high at 88%, so the schema already documents most parameters. The description adds one valuable clarification: content for guide_post_compact carries host compaction text, but it does not add meaning for the other parameters or nested summary fields.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific verb and resource: 'Record verified root lifecycle events,' and explicitly distinguishes this from 'ordinary saves,' which helps separate it from mem_save. It is less clear about what a root lifecycle event is, but the operation enum and sibling contrast keep the purpose identifiable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It gives one useful exclusion, 'never ordinary saves,' which points agents away from mem_save for normal persistence. However, it does not explain when to use enroll, recover, capture_root, checkpoint_pre_compact, guide_post_compact, or finalize, so operation selection remains implied by the schema rather than guided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
6 tool updates
v0.6.2- Changed
mem_context2 fields changed- removed
Input schema / properties / project_key / minLengthRemoved value: -1 - removed
Input schema / properties / root_session_key / minLengthRemoved value: -1
- Changed
mem_get1 field changed- changed
Input schema / properties / history / descriptionPrevious value: -"true: include predecessor lineage; evidence has no lineage."New value: +"Include predecessors; evidence has none."
- Changed
mem_project20 fields changed- removed
Input schema / properties / action / descriptionRemoved value: -"Project view." - changed
Input schema / properties / budget_chars / descriptionPrevious value: -"Aggregate character budget for timeline/briefing/summaries/observations; timeline clamps positive integers to 1024-20000."New value: +"Aggregate chars; timeline: positive integer clamped to 1024-20000." - changed
Input schema / properties / cursor / descriptionPrevious value: -"Timeline nextCursor, unchanged; keep project_key, since, until unchanged."New value: +"Unchanged nextCursor; preserve project_key, since, until." - removed
Input schema / properties / cursor / maxLengthRemoved value: -4096 - removed
Input schema / properties / cursor / minLengthRemoved value: -1 - removed
Input schema / properties / harness / descriptionRemoved value: -"Native harness for session filter." - removed
Input schema / properties / id / descriptionRemoved value: -"Prior-result record id." - removed
Input schema / properties / id / minLengthRemoved value: -1 - changed
Input schema / properties / limit / descriptionPrevious value: -"Maximum items for timeline/observations."New value: +"Timeline/observations: 1-100 items." - removed
Input schema / properties / limit / exclusiveMinimumRemoved value: -0 - removed
Input schema / properties / limit / maximumRemoved value: -100 - removed
Input schema / properties / project_key / minLengthRemoved value: -1 - changed
Input schema / properties / root_session_key / descriptionPrevious value: -"Verified session filter; pair with harness or omit both."New value: +"Session filter; pair with harness." - removed
Input schema / properties / root_session_key / minLengthRemoved value: -1 - changed
Input schema / properties / since / descriptionPrevious value: -"Timeline inclusive lower ISO-8601 instant; <= until."New value: +"Inclusive lower ISO-8601 instant; <= until." - removed
Input schema / properties / since / minLengthRemoved value: -1 - removed
Input schema / properties / state / descriptionRemoved value: -"Observations state filter." - changed
Input schema / properties / temporal / descriptionPrevious value: -"Summaries/observations only: current=latest summaries/correction leaves; history=older summaries/observation predecessors."New value: +"Summaries/observations: current=latest summaries/correction leaves; history=predecessors." - changed
Input schema / properties / until / descriptionPrevious value: -"Timeline inclusive upper ISO-8601 instant."New value: +"Inclusive upper ISO-8601 instant." - removed
Input schema / properties / until / minLengthRemoved value: -1
- Changed
mem_recall4 fields changed- changed
Input schema / properties / mode / descriptionPrevious value: -"compact (default): bounded snippets; context: selected content."New value: +"compact (default): snippets; context: selected content." - removed
Input schema / properties / project_key / minLengthRemoved value: -1 - removed
Input schema / properties / query / minLengthRemoved value: -1 - changed
Input schema / properties / temporal / descriptionPrevious value: -"current (default): current guidance; history: historical records."New value: +"current (default) or historical guidance."
- Changed
mem_save58 fields changed- removed
Input schema / properties / evidence / anyOfRemoved value: -[ - { - "additionalProperties": false, - "properties": { - "content": { - "minLength": 1, - "type": "string" - }, - "kind": { - "enum": [ - "root_prompt", - "explicit_save", - "checkpoint", - "handoff", - "legacy_prompt", - "legacy_observation", - "session_summary" - ], - "type": "string" - }, - "source_ref": { - "type": "string" - } - }, - "required": [ - "kind", - "content" - ], - "type": "object" - }, - { - "additionalProperties": false, - "properties": { - "content": { - "minLength": 1, - "type": "string" - }, - "kind": { - "const": "explicit_save", - "type": "string" - }, - "metadata": { - "additionalProperties": false, - "properties": { - "observation_validation": { - "additionalProperties": false, - "properties": { - "method": { - "maxLength": 500, - "minLength": 1, - "type": "string" - }, - "observation_id": { - "maxLength": 200, - "minLength": 1, - "type": "string" - }, - "result": { - "enum": [ - "passed", - "failed" - ], - "type": "string" - } - }, - "required": [ - "observation_id", - "result", - "method" - ], - "type": "object" - } - }, - "required": [ - "observation_validation" - ], - "type": "object" - }, - "source_ref": { - "type": "string" - } - }, - "required": [ - "kind", - "content", - "metadata" - ], - "type": "object" - }, - { - "additionalProperties": false, - "properties": { - "content": { - "minLength": 1, - "type": "string" - }, - "kind": { - "const": "handoff", - "type": "string" - }, - "metadata": { - "additionalProperties": false, - "properties": { - "observation_review_attestation": { - "additionalProperties": false, - "properties": { - "method": { - "maxLength": 500, - "minLength": 1, - "type": "string" - }, - "observation_id": { - "maxLength": 200, - "minLength": 1, - "type": "string" - }, - "reviewer": { - "maxLength": 200, - "minLength": 1, - "type": "string" - }, - "verdict": { - "enum": [ - "accepted", - "rejected" - ], - "type": "string" - } - }, - "required": [ - "observation_id", - "verdict", - "reviewer", - "method" - ], - "type": "object" - } - }, - "required": [ - "observation_review_attestation" - ], - "type": "object" - }, - "source_ref": { - "type": "string" - } - }, - "required": [ - "kind", - "content", - "metadata" - ], - "type": "object" - } -] - added
Input schema / properties / evidence / propertiesAdded value: +{ + "content": { + "type": "string" + }, + "kind": { + "enum": [ + "root_prompt", + "explicit_save", + "checkpoint", + "handoff", + "legacy_prompt", + "legacy_observation", + "session_summary" + ], + "type": "string" + }, + "metadata": { + "properties": { + "observation_review_attestation": { + "properties": { + "method": { + "type": "string" + }, + "observation_id": { + "type": "string" + }, + "reviewer": { + "type": "string" + }, + "verdict": { + "enum": [ + "accepted", + "rejected" + ], + "type": "string" + } + }, + "required": [ + "observation_id", + "verdict", + "reviewer", + "method" + ], + "type": "object" + }, + "observation_validation": { + "properties": { + "method": { + "type": "string" + }, + "observation_id": { + "type": "string" + }, + "result": { + "enum": [ + "passed", + "failed" + ], + "type": "string" + } + }, + "required": [ + "observation_id", + "result", + "method" + ], + "type": "object" + } + }, + "type": "object" + }, + "source_ref": { + "type": "string" + } +} - added
Input schema / properties / evidence / requiredAdded value: +[ + "kind", + "content" +] - added
Input schema / properties / evidence / typeAdded value: +"object" - removed
Input schema / properties / memory / additionalPropertiesRemoved value: -false - removed
Input schema / properties / memory / properties / content / minLengthRemoved value: -1 - removed
Input schema / properties / memory / properties / title / minLengthRemoved value: -1 - removed
Input schema / properties / observation / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation / properties / claim / maxLengthRemoved value: -4000 - removed
Input schema / properties / observation / properties / claim / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / concepts / items / maxLengthRemoved value: -200 - removed
Input schema / properties / observation / properties / concepts / items / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / concepts / maxItemsRemoved value: -16 - removed
Input schema / properties / observation / properties / coverage / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation / properties / coverage / properties / from_sequence / exclusiveMinimumRemoved value: -0 - removed
Input schema / properties / observation / properties / coverage / properties / from_sequence / maximumRemoved value: -9007199254740991 - removed
Input schema / properties / observation / properties / coverage / properties / to_sequence / exclusiveMinimumRemoved value: -0 - removed
Input schema / properties / observation / properties / coverage / properties / to_sequence / maximumRemoved value: -9007199254740991 - removed
Input schema / properties / observation / properties / files / items / maxLengthRemoved value: -500 - removed
Input schema / properties / observation / properties / files / items / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / files / maxItemsRemoved value: -16 - removed
Input schema / properties / observation / properties / generator / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation / properties / generator / properties / config_hash / patternRemoved value: -"^[0-9a-f]{64}$" - removed
Input schema / properties / observation / properties / generator / properties / name / maxLengthRemoved value: -200 - removed
Input schema / properties / observation / properties / generator / properties / name / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / generator / properties / version / maxLengthRemoved value: -200 - removed
Input schema / properties / observation / properties / generator / properties / version / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / predecessor_id / maxLengthRemoved value: -200 - removed
Input schema / properties / observation / properties / predecessor_id / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / proposed_memory / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation / properties / proposed_memory / properties / content / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / proposed_memory / properties / title / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / support_ids / items / maxLengthRemoved value: -200 - removed
Input schema / properties / observation / properties / support_ids / items / minLengthRemoved value: -1 - removed
Input schema / properties / observation / properties / support_ids / maxItemsRemoved value: -16 - removed
Input schema / properties / observation / properties / support_ids / minItemsRemoved value: -1 - removed
Input schema / properties / observation / properties / title / maxLengthRemoved value: -500 - removed
Input schema / properties / observation / properties / title / minLengthRemoved value: -1 - removed
Input schema / properties / observation_promotion / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation_promotion / properties / observation_id / maxLengthRemoved value: -200 - removed
Input schema / properties / observation_promotion / properties / observation_id / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation_review / properties / observation_id / maxLengthRemoved value: -200 - removed
Input schema / properties / observation_review / properties / observation_id / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / properties / policy / additionalPropertiesRemoved value: -false - removed
Input schema / properties / observation_review / properties / policy / properties / id / maxLengthRemoved value: -200 - removed
Input schema / properties / observation_review / properties / policy / properties / id / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / properties / policy / properties / version / maxLengthRemoved value: -200 - removed
Input schema / properties / observation_review / properties / policy / properties / version / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / properties / reason / maxLengthRemoved value: -1000 - removed
Input schema / properties / observation_review / properties / reason / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / properties / support_ids / items / maxLengthRemoved value: -200 - removed
Input schema / properties / observation_review / properties / support_ids / items / minLengthRemoved value: -1 - removed
Input schema / properties / observation_review / properties / support_ids / maxItemsRemoved value: -16 - removed
Input schema / properties / observation_review / properties / support_ids / minItemsRemoved value: -1 - removed
Input schema / properties / project_key / minLengthRemoved value: -1 - changed
Input schema / properties / project_name / descriptionPrevious value: -"Display/creation metadata only, not identity; prefer the persisted name from lifecycle/project output."New value: +"Display/creation name, not identity; prefer persisted lifecycle/project name." - removed
Input schema / properties / project_name / minLengthRemoved value: -1
- Changed
mem_session27 fields changed- removed
Input schema / properties / event_key / minLengthRemoved value: -1 - removed
Input schema / properties / project_key / minLengthRemoved value: -1 - changed
Input schema / properties / project_name / descriptionPrevious value: -"Display/creation metadata only, not identity; prefer the persisted name from lifecycle/project output."New value: +"Display/creation name, not identity; prefer persisted lifecycle/project name." - removed
Input schema / properties / project_name / minLengthRemoved value: -1 - removed
Input schema / properties / root_session_key / minLengthRemoved value: -1 - removed
Input schema / properties / summary / additionalPropertiesRemoved value: -false - removed
Input schema / properties / summary / properties / claims / items / additionalPropertiesRemoved value: -false - removed
Input schema / properties / summary / properties / claims / items / properties / content / minLengthRemoved value: -1 - changed
Input schema / properties / summary / properties / claims / items / properties / support_ids / descriptionPrevious value: -"Distinct evidence ids inside coverage from this project_key + root_session_key/harness session; never memory, summary, or observation ids."New value: +"Distinct same-session evidence ids within coverage; no memory/summary/observation ids." - removed
Input schema / properties / summary / properties / claims / items / properties / support_ids / items / maxLengthRemoved value: -200 - removed
Input schema / properties / summary / properties / claims / items / properties / support_ids / items / minLengthRemoved value: -1 - removed
Input schema / properties / summary / properties / claims / items / properties / support_ids / maxItemsRemoved value: -16 - removed
Input schema / properties / summary / properties / claims / items / properties / support_ids / minItemsRemoved value: -1 - removed
Input schema / properties / summary / properties / claims / maxItemsRemoved value: -32 - removed
Input schema / properties / summary / properties / claims / minItemsRemoved value: -1 - removed
Input schema / properties / summary / properties / coverage / additionalPropertiesRemoved value: -false - removed
Input schema / properties / summary / properties / coverage / properties / from_sequence / exclusiveMinimumRemoved value: -0 - removed
Input schema / properties / summary / properties / coverage / properties / from_sequence / maximumRemoved value: -9007199254740991 - changed
Input schema / properties / summary / properties / coverage / properties / to_sequence / descriptionPrevious value: -"Last sequence: >= from_sequence and >= the current summary's ending sequence for this session."New value: +">= from_sequence and the previous summary end." - removed
Input schema / properties / summary / properties / coverage / properties / to_sequence / exclusiveMinimumRemoved value: -0 - removed
Input schema / properties / summary / properties / coverage / properties / to_sequence / maximumRemoved value: -9007199254740991 - removed
Input schema / properties / summary / properties / generator / additionalPropertiesRemoved value: -false - removed
Input schema / properties / summary / properties / generator / properties / config_hash / patternRemoved value: -"^[0-9a-f]{64}$" - removed
Input schema / properties / summary / properties / generator / properties / name / maxLengthRemoved value: -200 - removed
Input schema / properties / summary / properties / generator / properties / name / minLengthRemoved value: -1 - removed
Input schema / properties / summary / properties / generator / properties / version / maxLengthRemoved value: -200 - removed
Input schema / properties / summary / properties / generator / properties / version / minLengthRemoved value: -1
6 tool updates
v0.6.1- Changed
mem_context3 fields changed- changed
Input schema / properties / harness / descriptionPrevious value: -"Native harness for root_session_key; supply both for session context, or omit both for project context."New value: +"Verified native harness." - changed
Input schema / properties / project_key / descriptionPrevious value: -"Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content."New value: +"Verified project_key, copied verbatim; never derived." - changed
Input schema / properties / root_session_key / descriptionPrevious value: -"Optional verified root session key; supply with harness for session context, or omit both for project context."New value: +"Verified root session key."
- Changed
mem_get2 fields changed- changed
Input schema / properties / history / descriptionPrevious value: -"Set true to expand predecessor lineage for memory, summary, or observation records; evidence ids have no lineage."New value: +"true: include predecessor lineage; evidence has no lineage." - changed
Input schema / properties / id / descriptionPrevious value: -"Memory, summary, observation, or evidence id returned by a prior tool result."New value: +"Memory, summary, observation, or evidence id."
- Changed
mem_project12 fields changed- changed
Input schema / properties / action / descriptionPrevious value: -"Inspect project views without mutation. action=\"list\": no fields required. action=\"timeline\": project_key required; optionally since, until, cursor, limit, and budget_chars. action=\"briefing\", \"summaries\", or \"observations\": project_key required; optionally supply both root_session_key and harness to select that session's summaries/observations (briefing still includes project-wide memories). action=\"history\": id required from a prior result. temporal filters summaries/observations, not timeline. list returns at most 256 exact aliases per project plus aliasCount and aliasesTruncated metadata."New value: +"Project view." - changed
Input schema / properties / budget_chars / descriptionPrevious value: -"Optional aggregate character budget for timeline, briefing, summaries, or observations; timeline clamps a positive integer to 1024-20000 characters."New value: +"Aggregate character budget for timeline/briefing/summaries/observations; timeline clamps positive integers to 1024-20000." - changed
Input schema / properties / cursor / descriptionPrevious value: -"Unchanged nextCursor from a prior mem_project action=\"timeline\" result; keep project_key, since, and until unchanged."New value: +"Timeline nextCursor, unchanged; keep project_key, since, until unchanged." - changed
Input schema / properties / harness / descriptionPrevious value: -"Native harness for the optional root_session_key filter; supply both or omit both. Not accepted for action=\"timeline\"."New value: +"Native harness for session filter." - changed
Input schema / properties / id / descriptionPrevious value: -"Required for action=\"history\"; send a record id returned by a prior tool result."New value: +"Prior-result record id." - changed
Input schema / properties / limit / descriptionPrevious value: -"Maximum items (1-100) for action=\"timeline\" or \"observations\"."New value: +"Maximum items for timeline/observations." - changed
Input schema / properties / project_key / descriptionPrevious value: -"Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content. Required for action=\"timeline\", \"briefing\", \"summaries\", or \"observations\"; not required for \"list\" or \"history\"."New value: +"Verified project_key, copied verbatim; never derived." - changed
Input schema / properties / root_session_key / descriptionPrevious value: -"Optional session filter for action=\"briefing\", \"summaries\", or \"observations\"; supply with harness or omit both."New value: +"Verified session filter; pair with harness or omit both." - changed
Input schema / properties / since / descriptionPrevious value: -"Optional inclusive lower ISO-8601 instant for action=\"timeline\" only; must be <= until."New value: +"Timeline inclusive lower ISO-8601 instant; <= until." - changed
Input schema / properties / state / descriptionPrevious value: -"Optional state filter for action=\"observations\"."New value: +"Observations state filter." - changed
Input schema / properties / temporal / descriptionPrevious value: -"Filter summaries/observations: current selects current summary versions or observation correction-chain leaves; history includes older summary versions or selects observation predecessors. Not accepted for action=\"timeline\"."New value: +"Summaries/observations only: current=latest summaries/correction leaves; history=older summaries/observation predecessors." - changed
Input schema / properties / until / descriptionPrevious value: -"Optional inclusive upper ISO-8601 instant for action=\"timeline\" only; must be >= since."New value: +"Timeline inclusive upper ISO-8601 instant."
- Changed
mem_recall4 fields changed- changed
Input schema / properties / mode / descriptionPrevious value: -"compact (default) returns bounded snippets; context adds selected memory content."New value: +"compact (default): bounded snippets; context: selected content." - changed
Input schema / properties / project_key / descriptionPrevious value: -"Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content."New value: +"Verified project_key, copied verbatim; never derived." - changed
Input schema / properties / query / descriptionPrevious value: -"Non-empty search string for promoted project memory."New value: +"Non-empty task search terms." - changed
Input schema / properties / temporal / descriptionPrevious value: -"current (default) searches current guidance; history includes historical records."New value: +"current (default): current guidance; history: historical records."
- Changed
mem_save10 fields changed- changed
Input schema / properties / event_key / descriptionPrevious value: -"Stable event key for idempotency; required for observation, observation_review, observation_promotion, and evidence with metadata."New value: +"Stable idempotency key." - changed
Input schema / properties / evidence / descriptionPrevious value: -"Direct evidence { kind, content }; optionally add memory for a promoted save. Structured metadata forbids memory and requires event_key, root_session_key, and harness."New value: +"Compact factual evidence; metadata requires event_key + session pair and forbids memory." - changed
Input schema / properties / harness / descriptionPrevious value: -"Native harness for root_session_key; supply both or omit both. Required wherever root_session_key is required."New value: +"Verified native harness." - changed
Input schema / properties / memory / descriptionPrevious value: -"Promoted memory for direct evidence only; requires evidence without metadata. Do not combine with observation, observation_review, or observation_promotion."New value: +"Direct promoted memory; requires evidence without metadata." - changed
Input schema / properties / observation / descriptionPrevious value: -"Submit a supported candidate without promotion; requires event_key. Session scope also requires root_session_key, harness, and coverage."New value: +"Supported candidate, not promoted; requires event_key. Session scope needs session pair + coverage; project scope forbids coverage." - changed
Input schema / properties / observation_promotion / descriptionPrevious value: -"Promote one accepted observation without new prose; requires event_key, root_session_key, and harness. Send no other operation branch or memory."New value: +"Promote accepted observation without new prose; requires event_key + session pair." - changed
Input schema / properties / observation_review / descriptionPrevious value: -"Review one supported observation; requires event_key, root_session_key, and harness. Send no other operation branch or memory."New value: +"Review supported observation; requires event_key + session pair." - changed
Input schema / properties / project_key / descriptionPrevious value: -"Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content."New value: +"Copy opaque project_key verbatim from verified native identity; never derive from names, paths, remotes, branches, worktrees, host IDs, listings, or recalled content." - changed
Input schema / properties / project_name / descriptionPrevious value: -"Creation/display metadata only; never participates in project identity equality. Prefer the database-persisted name returned by lifecycle or project output."New value: +"Display/creation metadata only, not identity; prefer the persisted name from lifecycle/project output." - changed
Input schema / properties / root_session_key / descriptionPrevious value: -"Verified root session key; supply with harness or omit both. Required for session-scoped observations, observation_review, observation_promotion, and evidence with metadata; blank counts as absent."New value: +"Verified root session key; blank is absent."
- Changed
mem_session15 fields changed- changed
Input schema / properties / event_key / descriptionPrevious value: -"Stable lifecycle event key; retries for the same operation must resend identical content/summary."New value: +"Stable lifecycle key; retries must resend identical content/summary." - changed
Input schema / properties / harness / descriptionPrevious value: -"Verified native harness for root_session_key in project_key."New value: +"Verified native harness." - changed
Input schema / properties / operation / descriptionPrevious value: -"Lifecycle operation; summary is allowed only for checkpoint_pre_compact (checkpoint) or finalize (final)."New value: +"Root lifecycle operation." - changed
Input schema / properties / project_key / descriptionPrevious value: -"Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content."New value: +"Verified project_key, copied verbatim; never derived." - changed
Input schema / properties / project_name / descriptionPrevious value: -"Creation/display metadata only; never participates in project identity equality. Prefer the database-persisted name returned by lifecycle or project output."New value: +"Display/creation metadata only, not identity; prefer the persisted name from lifecycle/project output." - changed
Input schema / properties / root_session_key / descriptionPrevious value: -"Stable verified root session key; required with harness for every operation."New value: +"Stable verified root session key." - changed
Input schema / properties / summary / descriptionPrevious value: -"Optional only for operation=\"checkpoint_pre_compact\" (kind=\"checkpoint\") or operation=\"finalize\" (kind=\"final\"). coverage starts at 1 and to_sequence must advance the current summary for this session and kind. support_ids must be evidence ids from this project_key + root_session_key/harness session inside coverage. Canonical submission limit: 20000 UTF-16 units."New value: +"Session summary." - changed
Input schema / properties / summary / properties / claims / descriptionPrevious value: -"1-32 atomic supported claims; keep the canonical summary within 20000 UTF-16 units."New value: +"Supported claims; canonical summary <=20000 UTF-16 units." - changed
Input schema / properties / summary / properties / claims / items / properties / content / descriptionPrevious value: -"Atomic claim content; at most 2000 code points after privacy filtering."New value: +"Atomic claim; <=2000 code points after privacy filtering." - changed
Input schema / properties / summary / properties / claims / items / properties / support_ids / descriptionPrevious value: -"1-16 distinct evidence ids from this project_key + root_session_key/harness session inside summary.coverage; never memory, summary, or observation ids."New value: +"Distinct evidence ids inside coverage from this project_key + root_session_key/harness session; never memory, summary, or observation ids." - changed
Input schema / properties / summary / properties / coverage / descriptionPrevious value: -"Inclusive session evidence coverage; start at 1 and advance to_sequence for this session and summary kind."New value: +"Inclusive session evidence range." - changed
Input schema / properties / summary / properties / coverage / properties / from_sequence / descriptionPrevious value: -"Inclusive first session event sequence; must start at 1."New value: +"First event sequence: 1." - changed
Input schema / properties / summary / properties / coverage / properties / to_sequence / descriptionPrevious value: -"Inclusive last session event sequence; must be >= from_sequence and exceed the current summary ending sequence for this session and kind."New value: +"Last sequence: >= from_sequence and >= the current summary's ending sequence for this session." - removed
Input schema / properties / summary / properties / kindRemoved value: -{ - "description": "Send checkpoint for operation=\"checkpoint_pre_compact\" or final for operation=\"finalize\".", - "enum": [ - "checkpoint", - "final" - ], - "type": "string" -} - changed
Input schema / properties / summary / requiredPrevious value: -[ - "kind", - "coverage", - "generator", - "claims" -]New value: +[ + "coverage", + "generator", + "claims" +]
6 tool updates
v0.5.6- Changed
mem_context13 fields changed- added
Input schema / properties / budget_charsAdded value: +{ + "type": "number" +} - added
Input schema / properties / correlation_idAdded value: +{ + "type": "string" +} - added
Input schema / properties / finalize_answerAdded value: +{ + "type": "boolean" +} - added
Input schema / properties / harnessAdded value: +{ + "description": "Native harness for root_session_key; supply both for session context, or omit both for project context.", + "enum": [ + "opencode", + "codex", + "claude", + "pi", + "mcp", + "cli", + "import" + ], + "type": "string" +} - removed
Input schema / properties / limitRemoved value: -{ - "description": "Number of observations to retrieve (default: 20)", - "type": "number" -} - removed
Input schema / properties / max_charsRemoved value: -{ - "description": "Output character budget; 0 disables the context cap", - "minimum": 0, - "type": "number" -} - removed
Input schema / properties / projectRemoved value: -{ - "description": "Filter by project name", - "type": "string" -} - added
Input schema / properties / project_keyAdded value: +{ + "description": "Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content.", + "minLength": 1, + "type": "string" +} - removed
Input schema / properties / recall_queryRemoved value: -{ - "description": "Optional query to append fused recall evidence without changing base context sections", - "type": "string" -} - added
Input schema / properties / root_session_keyAdded value: +{ + "description": "Optional verified root session key; supply with harness for session context, or omit both for project context.", + "minLength": 1, + "type": "string" +} - removed
Input schema / properties / scopeRemoved value: -{ - "description": "Filter by scope", - "enum": [ - "project", - "personal" - ], - "type": "string" -} - removed
Input schema / properties / session_idRemoved value: -{ - "description": "Filter to a specific session", - "type": "string" -} - added
Input schema / requiredAdded value: +[ + "project_key" +]
- Changed
mem_get10 fields changed- removed
Input schema / properties / afterRemoved value: -{ - "description": "Timeline observations after the focus item (default: 5)", - "maximum": 20, - "minimum": 0, - "type": "number" -} - removed
Input schema / properties / beforeRemoved value: -{ - "description": "Timeline observations before the focus item (default: 5)", - "maximum": 20, - "minimum": 0, - "type": "number" -} - added
Input schema / properties / correlation_idAdded value: +{ + "type": "string" +} - added
Input schema / properties / historyAdded value: +{ + "description": "Set true to expand predecessor lineage for memory, summary, or observation records; evidence ids have no lineage.", + "type": "boolean" +} - changed
Input schema / properties / id / descriptionPrevious value: -"Record ID to retrieve, interpreted according to kind"New value: +"Memory, summary, observation, or evidence id returned by a prior tool result." - changed
Input schema / properties / id / typePrevious value: -"number"New value: +"string" - removed
Input schema / properties / include_timelineRemoved value: -{ - "description": "Include surrounding observations in the same session", - "type": "boolean" -} - removed
Input schema / properties / kindRemoved value: -{ - "description": "Memory kind to retrieve (defaults to observation)", - "enum": [ - "observation", - "prompt" - ], - "type": "string" -} - removed
Input schema / properties / max_lengthRemoved value: -{ - "description": "Max characters to return (default: 50000)", - "minimum": 100, - "type": "number" -} - removed
Input schema / properties / offsetRemoved value: -{ - "description": "Character offset for large content (default: 0)", - "minimum": 0, - "type": "number" -}
- Changed
mem_project26 fields changed- changed
Input schema / properties / action / descriptionPrevious value: -"Project view to return"New value: +"Inspect project views without mutation. action=\"list\": no fields required. action=\"timeline\": project_key required; optionally since, until, cursor, limit, and budget_chars. action=\"briefing\", \"summaries\", or \"observations\": project_key required; optionally supply both root_session_key and harness to select that session's summaries/observations (briefing still includes project-wide memories). action=\"history\": id required from a prior result. temporal filters summaries/observations, not timeline. list returns at most 256 exact aliases per project plus aliasCount and aliasesTruncated metadata." - changed
Input schema / properties / action / enumPrevious value: -[ - "list", - "summary", - "graph", - "topics", - "topic", - "health" -]New value: +[ + "list", + "timeline", + "briefing", + "history", + "summaries", + "observations" +] - added
Input schema / properties / budget_charsAdded value: +{ + "description": "Optional aggregate character budget for timeline, briefing, summaries, or observations; timeline clamps a positive integer to 1024-20000 characters.", + "type": "number" +} - removed
Input schema / properties / continuationRemoved value: -{ - "description": "Opaque continuation token returned by graph navigation views", - "type": "string" -} - added
Input schema / properties / cursorAdded value: +{ + "description": "Unchanged nextCursor from a prior mem_project action=\"timeline\" result; keep project_key, since, and until unchanged.", + "maxLength": 4096, + "minLength": 1, + "type": "string" +} - removed
Input schema / properties / focus_node_idRemoved value: -{ - "description": "Graph focus node id for navigation=neighborhood, currently obs:<id>", - "type": "string" -} - added
Input schema / properties / harnessAdded value: +{ + "description": "Native harness for the optional root_session_key filter; supply both or omit both. Not accepted for action=\"timeline\".", + "enum": [ + "opencode", + "codex", + "claude", + "pi", + "mcp", + "cli", + "import" + ], + "type": "string" +} - added
Input schema / properties / idAdded value: +{ + "description": "Required for action=\"history\"; send a record id returned by a prior tool result.", + "minLength": 1, + "type": "string" +} - removed
Input schema / properties / include_supersededRemoved value: -{ - "description": "Explicit history opt-in; honored by navigation=superseded only", - "type": "boolean" -} - changed
Input schema / properties / limit / descriptionPrevious value: -"Maximum items to return"New value: +"Maximum items (1-100) for action=\"timeline\" or \"observations\"." - added
Input schema / properties / limit / exclusiveMinimumAdded value: +0 - changed
Input schema / properties / limit / maximumPrevious value: -500New value: +100 - removed
Input schema / properties / limit / minimumRemoved value: -1 - changed
Input schema / properties / limit / typePrevious value: -"number"New value: +"integer" - removed
Input schema / properties / max_charsRemoved value: -{ - "description": "Response character budget; 0 is supported for action=summary only", - "maximum": 20000, - "minimum": 0, - "type": "integer" -} - removed
Input schema / properties / navigationRemoved value: -{ - "description": "Graph navigation mode for action=graph; defaults to ledger", - "enum": [ - "ledger", - "neighborhood", - "lineage", - "community", - "superseded" - ], - "type": "string" -} - removed
Input schema / properties / observation_idRemoved value: -{ - "description": "Observation id for lineage or superseded graph navigation", - "exclusiveMinimum": 0, - "maximum": 9007199254740991, - "type": "integer" -} - removed
Input schema / properties / projectRemoved value: -{ - "description": "Project name. Required except action=list and optional for action=topics or health", - "type": "string" -} - added
Input schema / properties / project_keyAdded value: +{ + "description": "Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content. Required for action=\"timeline\", \"briefing\", \"summaries\", or \"observations\"; not required for \"list\" or \"history\".", + "minLength": 1, + "type": "string" +} - removed
Input schema / properties / relationRemoved value: -{ - "description": "Graph relation filter for action=graph", - "enum": [ - "HAS_TYPE", - "IN_PROJECT", - "HAS_TOPIC_KEY", - "HAS_WHAT", - "HAS_WHY", - "HAS_WHERE", - "HAS_LEARNED" - ], - "type": "string" -} - added
Input schema / properties / root_session_keyAdded value: +{ + "description": "Optional session filter for action=\"briefing\", \"summaries\", or \"observations\"; supply with harness or omit both.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / sinceAdded value: +{ + "description": "Optional inclusive lower ISO-8601 instant for action=\"timeline\" only; must be <= until.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / stateAdded value: +{ + "description": "Optional state filter for action=\"observations\".", + "enum": [ + "pending", + "accepted", + "rejected", + "promoted" + ], + "type": "string" +} - added
Input schema / properties / temporalAdded value: +{ + "description": "Filter summaries/observations: current selects current summary versions or observation correction-chain leaves; history includes older summary versions or selects observation predecessors. Not accepted for action=\"timeline\".", + "enum": [ + "current", + "history" + ], + "type": "string" +} - removed
Input schema / properties / topic_keyRemoved value: -{ - "description": "Topic key for action=topic or graph filtering", - "type": "string" -} - added
Input schema / properties / untilAdded value: +{ + "description": "Optional inclusive upper ISO-8601 instant for action=\"timeline\" only; must be >= since.", + "minLength": 1, + "type": "string" +}
- Changed
mem_recall20 fields changed- added
Input schema / properties / budget_charsAdded value: +{ + "type": "number" +} - added
Input schema / properties / correlation_idAdded value: +{ + "type": "string" +} - removed
Input schema / properties / debugRemoved value: -{ - "description": "Include retrieval defaults and semantic input sources", - "type": "boolean" -} - added
Input schema / properties / finalize_answerAdded value: +{ + "type": "boolean" +} - removed
Input schema / properties / hydeRemoved value: -{ - "description": "Request HyDE query expansion when configured", - "type": "boolean" -} - removed
Input schema / properties / limit / descriptionRemoved value: -"Maximum evidence items (default: 5)" - removed
Input schema / properties / limit / maximumRemoved value: -20 - removed
Input schema / properties / limit / minimumRemoved value: -1 - changed
Input schema / properties / mode / descriptionPrevious value: -"compact returns evidence lines; context includes retrieved text"New value: +"compact (default) returns bounded snippets; context adds selected memory content." - removed
Input schema / properties / projectRemoved value: -{ - "description": "Optional project filter", - "type": "string" -} - added
Input schema / properties / project_keyAdded value: +{ + "description": "Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content.", + "minLength": 1, + "type": "string" +} - changed
Input schema / properties / query / descriptionPrevious value: -"Recall/search query"New value: +"Non-empty search string for promoted project memory." - removed
Input schema / properties / scopeRemoved value: -{ - "description": "Optional scope filter", - "enum": [ - "project", - "personal" - ], - "type": "string" -} - removed
Input schema / properties / session_idRemoved value: -{ - "description": "Optional session filter", - "type": "string" -} - added
Input schema / properties / temporalAdded value: +{ + "description": "current (default) searches current guidance; history includes historical records.", + "enum": [ + "current", + "history" + ], + "type": "string" +} - removed
Input schema / properties / time_fromRemoved value: -{ - "description": "Optional inclusive created_at lower bound", - "type": "string" -} - removed
Input schema / properties / time_toRemoved value: -{ - "description": "Optional inclusive created_at upper bound", - "type": "string" -} - removed
Input schema / properties / topic_keyRemoved value: -{ - "description": "Optional exact topic_key filter", - "type": "string" -} - removed
Input schema / properties / typeRemoved value: -{ - "description": "Optional observation type filter", - "enum": [ - "decision", - "architecture", - "bugfix", - "pattern", - "config", - "discovery", - "learning", - "session_summary", - "manual" - ], - "type": "string" -} - changed
Input schema / requiredPrevious value: -[ - "query" -]New value: +[ + "project_key", + "query" +]
- Changed
mem_save19 fields changed- removed
Input schema / properties / contentRemoved value: -{ - "description": "Memory content, prompt text, session summary, or text containing a Key Learnings section", - "type": "string" -} - added
Input schema / properties / event_keyAdded value: +{ + "description": "Stable event key for idempotency; required for observation, observation_review, observation_promotion, and evidence with metadata.", + "type": "string" +} - added
Input schema / properties / evidenceAdded value: +{ + "anyOf": [ + { + "additionalProperties": false, + "properties": { + "content": { + "minLength": 1, + "type": "string" + }, + "kind": { + "enum": [ + "root_prompt", + "explicit_save", + "checkpoint", + "handoff", + "legacy_prompt", + "legacy_observation", + "session_summary" + ], + "type": "string" + }, + "source_ref": { + "type": "string" + } + }, + "required": [ + "kind", + "content" + ], + "type": "object" + }, + { + "additionalProperties": false, + "properties": { + "content": { + "minLength": 1, + "type": "string" + }, + "kind": { + "const": "explicit_save", + "type": "string" + }, + "metadata": { + "additionalProperties": false, + "properties": { + "observation_validation": { + "additionalProperties": false, + "properties": { + "method": { + "maxLength": 500, + "minLength": 1, + "type": "string" + }, + "observation_id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "result": { + "enum": [ + "passed", + "failed" + ], + "type": "string" + } + }, + "required": [ + "observation_id", + "result", + "method" + ], + "type": "object" + } + }, + "required": [ + "observation_validation" + ], + "type": "object" + }, + "source_ref": { + "type": "string" + } + }, + "required": [ + "kind", + "content", + "metadata" + ], + "type": "object" + }, + { + "additionalProperties": false, + "properties": { + "content": { + "minLength": 1, + "type": "string" + }, + "kind": { + "const": "handoff", + "type": "string" + }, + "metadata": { + "additionalProperties": false, + "properties": { + "observation_review_attestation": { + "additionalProperties": false, + "properties": { + "method": { + "maxLength": 500, + "minLength": 1, + "type": "string" + }, + "observation_id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "reviewer": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "verdict": { + "enum": [ + "accepted", + "rejected" + ], + "type": "string" + } + }, + "required": [ + "observation_id", + "verdict", + "reviewer", + "method" + ], + "type": "object" + } + }, + "required": [ + "observation_review_attestation" + ], + "type": "object" + }, + "source_ref": { + "type": "string" + } + }, + "required": [ + "kind", + "content", + "metadata" + ], + "type": "object" + } + ], + "description": "Direct evidence { kind, content }; optionally add memory for a promoted save. Structured metadata forbids memory and requires event_key, root_session_key, and harness." +} - added
Input schema / properties / harnessAdded value: +{ + "description": "Native harness for root_session_key; supply both or omit both. Required wherever root_session_key is required.", + "enum": [ + "opencode", + "codex", + "claude", + "pi", + "mcp", + "cli", + "import" + ], + "type": "string" +} - removed
Input schema / properties / kindRemoved value: -{ - "description": "Write mode. Defaults to observation", - "enum": [ - "observation", - "prompt", - "session_summary", - "passive_learnings" - ], - "type": "string" -} - added
Input schema / properties / memoryAdded value: +{ + "additionalProperties": false, + "description": "Promoted memory for direct evidence only; requires evidence without metadata. Do not combine with observation, observation_review, or observation_promotion.", + "properties": { + "content": { + "minLength": 1, + "type": "string" + }, + "kind": { + "enum": [ + "decision", + "convention", + "architecture", + "discovery", + "failure", + "project_structure", + "handoff", + "preference" + ], + "type": "string" + }, + "outcome": { + "enum": [ + "unknown", + "succeeded", + "failed", + "mixed" + ], + "type": "string" + }, + "supersedes_id": { + "type": "string" + }, + "title": { + "minLength": 1, + "type": "string" + }, + "topic_key": { + "type": "string" + } + }, + "required": [ + "kind", + "title", + "content" + ], + "type": "object" +} - added
Input schema / properties / observationAdded value: +{ + "additionalProperties": false, + "description": "Submit a supported candidate without promotion; requires event_key. Session scope also requires root_session_key, harness, and coverage.", + "properties": { + "claim": { + "maxLength": 4000, + "minLength": 1, + "type": "string" + }, + "concepts": { + "items": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "maxItems": 16, + "type": "array" + }, + "coverage": { + "additionalProperties": false, + "properties": { + "from_sequence": { + "exclusiveMinimum": 0, + "maximum": 9007199254740991, + "type": "integer" + }, + "to_sequence": { + "exclusiveMinimum": 0, + "maximum": 9007199254740991, + "type": "integer" + } + }, + "required": [ + "from_sequence", + "to_sequence" + ], + "type": "object" + }, + "files": { + "items": { + "maxLength": 500, + "minLength": 1, + "type": "string" + }, + "maxItems": 16, + "type": "array" + }, + "generator": { + "additionalProperties": false, + "properties": { + "config_hash": { + "pattern": "^[0-9a-f]{64}$", + "type": "string" + }, + "kind": { + "enum": [ + "root_agent", + "harness", + "model" + ], + "type": "string" + }, + "name": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "version": { + "maxLength": 200, + "minLength": 1, + "type": "string" + } + }, + "required": [ + "kind", + "name" + ], + "type": "object" + }, + "kind": { + "enum": [ + "decision", + "constraint", + "fact", + "procedure", + "result", + "failure", + "preference" + ], + "type": "string" + }, + "predecessor_id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "proposed_memory": { + "additionalProperties": false, + "properties": { + "content": { + "minLength": 1, + "type": "string" + }, + "kind": { + "enum": [ + "decision", + "convention", + "architecture", + "discovery", + "failure", + "project_structure", + "handoff", + "preference" + ], + "type": "string" + }, + "outcome": { + "enum": [ + "unknown", + "succeeded", + "failed", + "mixed" + ], + "type": "string" + }, + "title": { + "minLength": 1, + "type": "string" + }, + "topic_key": { + "type": "string" + } + }, + "required": [ + "kind", + "title", + "content" + ], + "type": "object" + }, + "scope": { + "enum": [ + "session", + "project" + ], + "type": "string" + }, + "support_ids": { + "items": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "maxItems": 16, + "minItems": 1, + "type": "array" + }, + "title": { + "maxLength": 500, + "minLength": 1, + "type": "string" + } + }, + "required": [ + "kind", + "scope", + "title", + "claim", + "proposed_memory", + "support_ids", + "generator" + ], + "type": "object" +} - added
Input schema / properties / observation_promotionAdded value: +{ + "additionalProperties": false, + "description": "Promote one accepted observation without new prose; requires event_key, root_session_key, and harness. Send no other operation branch or memory.", + "properties": { + "observation_id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + } + }, + "required": [ + "observation_id" + ], + "type": "object" +} - added
Input schema / properties / observation_reviewAdded value: +{ + "additionalProperties": false, + "description": "Review one supported observation; requires event_key, root_session_key, and harness. Send no other operation branch or memory.", + "properties": { + "basis": { + "enum": [ + "root_user_confirmed", + "observable_validation", + "independent_review" + ], + "type": "string" + }, + "observation_id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "policy": { + "additionalProperties": false, + "properties": { + "id": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "version": { + "maxLength": 200, + "minLength": 1, + "type": "string" + } + }, + "required": [ + "id", + "version" + ], + "type": "object" + }, + "reason": { + "maxLength": 1000, + "minLength": 1, + "type": "string" + }, + "support_ids": { + "items": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "maxItems": 16, + "minItems": 1, + "type": "array" + }, + "verdict": { + "enum": [ + "accepted", + "rejected" + ], + "type": "string" + } + }, + "required": [ + "observation_id", + "verdict", + "basis", + "policy", + "reason", + "support_ids" + ], + "type": "object" +} - removed
Input schema / properties / projectRemoved value: -{ - "description": "Project name", - "type": "string" -} - added
Input schema / properties / project_keyAdded value: +{ + "description": "Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / project_nameAdded value: +{ + "description": "Creation/display metadata only; never participates in project identity equality. Prefer the database-persisted name returned by lifecycle or project output.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / root_session_keyAdded value: +{ + "description": "Verified root session key; supply with harness or omit both. Required for session-scoped observations, observation_review, observation_promotion, and evidence with metadata; blank counts as absent.", + "type": "string" +} - removed
Input schema / properties / scopeRemoved value: -{ - "description": "Observation scope", - "enum": [ - "project", - "personal" - ], - "type": "string" -} - removed
Input schema / properties / session_idRemoved value: -{ - "description": "Session ID (default: manual-save-{project})", - "type": "string" -} - removed
Input schema / properties / titleRemoved value: -{ - "description": "Short searchable title. Required for kind=observation", - "type": "string" -} - removed
Input schema / properties / topic_keyRemoved value: -{ - "description": "Stable key for observation upserts", - "type": "string" -} - removed
Input schema / properties / typeRemoved value: -{ - "description": "Observation category for kind=observation", - "enum": [ - "decision", - "architecture", - "bugfix", - "pattern", - "config", - "discovery", - "learning", - "session_summary", - "manual" - ], - "type": "string" -} - changed
Input schema / requiredPrevious value: -[ - "content" -]New value: +[ + "project_key", + "project_name" +]
- Changed
mem_session17 fields changed- removed
Input schema / properties / actionRemoved value: -{ - "description": "Session action", - "enum": [ - "start", - "summary", - "checkpoint" - ], - "type": "string" -} - removed
Input schema / properties / content / descriptionRemoved value: -"Full session summary for action=summary" - removed
Input schema / properties / directoryRemoved value: -{ - "description": "Working directory for action=start", - "type": "string" -} - added
Input schema / properties / event_keyAdded value: +{ + "description": "Stable lifecycle event key; retries for the same operation must resend identical content/summary.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / harnessAdded value: +{ + "description": "Verified native harness for root_session_key in project_key.", + "enum": [ + "opencode", + "codex", + "claude", + "pi", + "mcp", + "cli", + "import" + ], + "type": "string" +} - removed
Input schema / properties / idRemoved value: -{ - "description": "Session ID. Required for action=start; defaults to manual-save-{project} for summary/checkpoint", - "type": "string" -} - added
Input schema / properties / operationAdded value: +{ + "description": "Lifecycle operation; summary is allowed only for checkpoint_pre_compact (checkpoint) or finalize (final).", + "enum": [ + "enroll", + "recover", + "capture_root", + "checkpoint_pre_compact", + "guide_post_compact", + "finalize" + ], + "type": "string" +} - removed
Input schema / properties / projectRemoved value: -{ - "description": "Project name", - "type": "string" -} - added
Input schema / properties / project_keyAdded value: +{ + "description": "Exact opaque project_key copied verbatim from verified native identity; never derive it from a display name, path hint, remote, branch, worktree name, host ID, listing, or recalled content.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / project_nameAdded value: +{ + "description": "Creation/display metadata only; never participates in project identity equality. Prefer the database-persisted name returned by lifecycle or project output.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / root_session_keyAdded value: +{ + "description": "Stable verified root session key; required with harness for every operation.", + "minLength": 1, + "type": "string" +} - added
Input schema / properties / summary / additionalPropertiesAdded value: +false - changed
Input schema / properties / summary / descriptionPrevious value: -"Short checkpoint summary for action=checkpoint"New value: +"Optional only for operation=\"checkpoint_pre_compact\" (kind=\"checkpoint\") or operation=\"finalize\" (kind=\"final\"). coverage starts at 1 and to_sequence must advance the current summary for this session and kind. support_ids must be evidence ids from this project_key + root_session_key/harness session inside coverage. Canonical submission limit: 20000 UTF-16 units." - added
Input schema / properties / summary / propertiesAdded value: +{ + "claims": { + "description": "1-32 atomic supported claims; keep the canonical summary within 20000 UTF-16 units.", + "items": { + "additionalProperties": false, + "properties": { + "content": { + "description": "Atomic claim content; at most 2000 code points after privacy filtering.", + "minLength": 1, + "type": "string" + }, + "kind": { + "enum": [ + "objective", + "completed", + "decision", + "changed_surface", + "verification", + "pending", + "blocker", + "next_action" + ], + "type": "string" + }, + "outcome": { + "enum": [ + "unknown", + "succeeded", + "failed", + "mixed" + ], + "type": "string" + }, + "support_ids": { + "description": "1-16 distinct evidence ids from this project_key + root_session_key/harness session inside summary.coverage; never memory, summary, or observation ids.", + "items": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "maxItems": 16, + "minItems": 1, + "type": "array" + } + }, + "required": [ + "kind", + "content", + "support_ids" + ], + "type": "object" + }, + "maxItems": 32, + "minItems": 1, + "type": "array" + }, + "coverage": { + "additionalProperties": false, + "description": "Inclusive session evidence coverage; start at 1 and advance to_sequence for this session and summary kind.", + "properties": { + "from_sequence": { + "description": "Inclusive first session event sequence; must start at 1.", + "exclusiveMinimum": 0, + "maximum": 9007199254740991, + "type": "integer" + }, + "to_sequence": { + "description": "Inclusive last session event sequence; must be >= from_sequence and exceed the current summary ending sequence for this session and kind.", + "exclusiveMinimum": 0, + "maximum": 9007199254740991, + "type": "integer" + } + }, + "required": [ + "from_sequence", + "to_sequence" + ], + "type": "object" + }, + "generator": { + "additionalProperties": false, + "properties": { + "config_hash": { + "pattern": "^[0-9a-f]{64}$", + "type": "string" + }, + "kind": { + "enum": [ + "root_agent", + "harness", + "model" + ], + "type": "string" + }, + "name": { + "maxLength": 200, + "minLength": 1, + "type": "string" + }, + "version": { + "maxLength": 200, + "minLength": 1, + "type": "string" + } + }, + "required": [ + "kind", + "name" + ], + "type": "object" + }, + "kind": { + "description": "Send checkpoint for operation=\"checkpoint_pre_compact\" or final for operation=\"finalize\".", + "enum": [ + "checkpoint", + "final" + ], + "type": "string" + } +} - added
Input schema / properties / summary / requiredAdded value: +[ + "kind", + "coverage", + "generator", + "claims" +] - changed
Input schema / properties / summary / typePrevious value: -"string"New value: +"object" - changed
Input schema / requiredPrevious value: -[ - "action", - "project" -]New value: +[ + "operation", + "harness", + "project_key", + "project_name", + "root_session_key", + "event_key" +]
5 tool updates
v0.4.13- Added
mem_context - Added
mem_get - Added
mem_project - Added
mem_save - Added
mem_session
5 tool updates
v0.4.1- Removed
mem_context - Removed
mem_get - Removed
mem_project - Removed
mem_save - Removed
mem_session
6 tool updates
v0.3.7- First observed
mem_context - First observed
mem_get - First observed
mem_project - First observed
mem_recall - First observed
mem_save - First observed
mem_session
TDQS
Scored across 6 tools
mem_save and mem_session both persist information, but mem_session is explicitly limited to lifecycle events. The read-side tools (mem_recall, mem_context, mem_project) have overlapping search/context/view roles, which could cause misselection, though descriptions provide clear trigger guidance.
All tools use a consistent mem_ prefix in snake_case, making names predictable and readable. Minor inconsistency exists because some names are verbs (save, get, recall) and others are nouns (context, project, session), but the pattern is still coherent.
Six tools is well within the ideal range for a focused memory server. Each tool maps to a distinct stage of the memory lifecycle—save, recall, expand, context building, project inspection, and session events—without obvious redundancy.
The surface covers saving, searching, retrieval, context building, project views, and session lifecycle. However, there is no explicit update or delete/forget operation for memories, and listing memories beyond search results or project views is limited, which may hinder correcting or pruning stored knowledge.
Maintenance
Related MCP Connectors
Persistent memory for AI agents. Search, store, and recall across sessions.
Shared memory for coding agents. Stop re-explaining your codebase every session.
Persistent memory for AI agents. Search and store durable facts, preferences and decisions.
Persistent cross-session memory shared by Codex, Claude Code, ChatGPT, and other AI agents.
Related MCP Servers
- AlicenseNot gradedqualityNot gradedmaintenanceProvides AI coding agents with persistent, long-term memory through local semantic search and SQLite storage. It enables agents to save and retrieve architectural decisions or project context across different conversation sessions without requiring cloud services.MIT
- AlicenseNot gradedqualityDmaintenanceProvides AI coding assistants with persistent project memory to retain architectural decisions, code patterns, and domain knowledge across sessions. It stores data locally in a SQLite database, allowing agents to remember, recall, and manage project-specific context using full-text search.0Apache 2.0
- AlicenseNot gradedqualityCmaintenanceGives AI coding agents persistent memory by storing observations, decisions, and learnings in a local SQLite database with vector search, full-text search, and a rules engine.4MIT
- AlicenseAqualityAmaintenanceLocal-first memory for AI coding agents using SQLite, enabling persistent recall of decisions, lessons, and project context without cloud or API keys.10157 PyPI286Apache 2.0