Skip to main content
Glama

Mnemoverse Memory

by io.github.mnemoverse
OverviewDiscussionsAdmin

memory_write

Store a long-term memory that persists across sessions AND across every AI tool the user has connected to Mnemoverse (Claude, ChatGPT, Cursor, VS Code) — write once, recall everywhere. Call this PROACTIVELY the moment the user states a preference, makes a decision, or you learn a durable fact (people, roles, project setup, a lesson). Don't wait to be asked. Never store passwords, API keys, payment data, MFA codes, government IDs, or health records; skip transient chatter that only matters this turn. Behavior: an importance gate may filter low-value writes, so the result tells you whether the memory was stored or filtered. Write content as a self-contained statement that still makes sense when recalled out of context.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
domainNoNamespace to organize memories (e.g. 'engineering', 'user:alice', 'project:acme'). Matched byte-for-byte — a leading space, a different case, or an invisible character opens a SEPARATE, permanent store, so reuse an exact name from memory_stats rather than retyping one. To write into a shared room, pass its address here instead (e.g. 'xroom:room_01ABC'). Find room addresses with memory_list_rooms.
contentYesThe memory to store as a self-contained statement, e.g. 'User prefers TypeScript strict mode' or 'Decided to deploy the API on Cloudflare Workers (2026-06)'.
conceptsNoKey concepts for linking related memories (e.g. ['deploy', 'friday', 'staging'])

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault
reasonNoThe memory service's own explanation of this outcome, quoted as sent — when stored is false this is the ONLY statement of WHY, e.g. "Below importance threshold (0.047 < 0.1)". Ordinary text is preserved exactly; only control, bidi, zero-width, and repeated-whitespace characters are normalized before display, and the value is capped at 400 characters. Absent when the service sent no explanation, or when nothing remains after that normalization.
storedYesWhether the memory passed the novelty gate and was stored.
memory_idYesIdentifier of the stored memory, or null when it was not stored.
importanceNoNovelty score for this write (0-1): how much it adds over the nearest memories already saved in the same domain. A first-generation metric UNDER ACTIVE DEVELOPMENT and known to be unreliable — the same content has measured ~0.08 in Russian against ~0.55 in English, so it under-reads non-English text. It is not a verdict on whether the memory was worth keeping. Absent when the service sent no score.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed7 schema fields changed
    • removedInput schema / additionalProperties
      Removed value: -false
    • addedInput schema / properties / concepts
      Added value: +{
+  "description": "Key concepts for linking related memories (e.g. ['deploy', 'friday', 'staging'])",
+  "items": {
+    "type": "string"
+  },
+  "maxItems": 256,
+  "type": "array"
+}
    • changedInput schema / properties / content / description
      Previous value: -"The text to remember (1-10,000 chars). Do not store passwords, API keys, payment data, MFA codes, government IDs, or health records."New value: +"The memory to store as a self-contained statement, e.g. 'User prefers TypeScript strict mode' or 'Decided to deploy the API on Cloudflare Workers (2026-06)'."
    • changedInput schema / properties / domain / description
      Previous value: -"Namespace: 'general', 'user:X', 'project:Z' (default 'general')."New value: +"Namespace to organize memories (e.g. 'engineering', 'user:alice', 'project:acme'). Matched byte-for-byte — a leading space, a different case, or an invisible character opens a SEPARATE, permanent store, so reuse an exact name from memory_stats rather than retyping one. To write into a shared room, pass its address here instead (e.g. 'xroom:room_01ABC'). Find room addresses with memory_list_rooms."
    • removedOutput schema / properties / memory_id / anyOf
      Removed value: -[
-  {
-    "format": "uuid",
-    "pattern": "^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12})$",
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]
    • addedOutput schema / properties / memory_id / type
      Added value: +[
+  "string",
+  "null"
+]
    • changedOutput schema / properties / reason / description
      Previous value: -"The memory service's own explanation of this outcome, quoted as sent — when stored is false this is the ONLY statement of WHY, e.g. \"Below importance threshold (0.047 < 0.1)\". Ordinary text is preserved exactly; only control, bidi, zero-width, and repeated-whitespace characters are normalized before display. Absent when the service sent no explanation."New value: +"The memory service's own explanation of this outcome, quoted as sent — when stored is false this is the ONLY statement of WHY, e.g. \"Below importance threshold (0.047 < 0.1)\". Ordinary text is preserved exactly; only control, bidi, zero-width, and repeated-whitespace characters are normalized before display, and the value is capped at 400 characters. Absent when the service sent no explanation, or when nothing remains after that normalization."
  2. First observed

TDQS

A5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the annotations, the description discloses that an importance gate may filter low-value writes and that the result indicates whether the memory was stored or filtered. It also warns about exact byte-for-byte domain matching creating separate permanent stores, which is critical behavioral context not available in the schema or annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is dense but every sentence carries operational value: outcome, triggers, exclusions, gate behavior, and content formatting guidance are all included. It is front-loaded with the core purpose, and the structured lists keep constraints scannable.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The description covers all essential dimensions for correct invocation: when to use, what not to store, how to format `content`, how `domain` matching behaves, how to reference rooms via memory_list_rooms, and what to expect from the filtering gate. With an output schema present and these details included, nothing essential is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Even though schema coverage is 100%, the description adds significant meaning: the `domain` parameter's exact-match and namespace behavior, the `content` requirement to be a self-contained statement with examples, and `concepts` as link keys for related memories. This goes well beyond the schema's basic field descriptions.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific action — 'Store a long-term memory' — and clearly defines its unique scope: persistence across sessions and across multiple connected AI tools. It distinguishes itself from sibling tools like memory_read and memory_list_rooms by being the write destination for durable knowledge.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives explicit when-to-use guidance: call proactively when the user states a preference, makes a decision, or reveals a durable fact, and don't wait to be asked. It also provides strong exclusions — never store passwords, API keys, payment data, MFA codes, government IDs, health records, or transient chatter — which helps an agent decide correctly.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Mnemoverse Memory toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources