Skip to main content
Glama
LiLara-AI

ShadowGraph

Official

shadowgraph_remember

Add or reconcile scoped memories by identity tuple, applying ADD, UPDATE, DELETE, or NOOP with full revision history. Solves duplicate and stale memory writes while keeping changes auditable.

Instructions

Add or reconcile one scoped memory, or apply an ADD/UPDATE/DELETE/NOOP plan, by identity tuple. Use shadowgraph_record_decision for a choice, shadowgraph_record_fact for an observation, shadowgraph_recall to read memory back. Identical content is a NOOP, new content supersedes and keeps history, DELETE invalidates; every call commits a revision.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
keyNoStable caller-chosen name within this scope and type, such as hotel-style. Reusing it reconciles that memory rather than adding a second one.
tagsNoFree-form labels stored with the memory. Part of the compared content, so changing them produces a new version.
textNoThe memory content itself. Stored verbatim, searched as content, and compared to decide ADD, UPDATE, or NOOP.
actorNoWho performed this write, such as an agent or person name. Stored for audit; never used to grant trust.
scopeNoScope selector. Omitted or partial fields mean explicit nulls, not "any": identity is the exact (project, userId, agentId, runId, memoryType, key) tuple, so a run-scoped memory never leaks into a user-only read.
clientNoWhich client software performed this write, such as the host application name.
projectNoProject namespace. Defaults to "default"; an empty string is rejected.
validToNoISO 8601 instant after which it stops being true, or null for open-ended. Must be later than validFrom.
metadataNoCaller-owned JSON object kept with the memory. Part of the compared content, so changing it produces a new version.
embeddingNoCaller-supplied vector for this text. Omit to use the configured embedding provider; with no provider the record is still stored and recall reports semantic.available=false instead of renaming lexical overlap as semantic.
sessionIdNoCaller-owned identifier that groups related writes in the audit trail.
validFromNoISO 8601 instant from which this memory is true in the modeled world. Defaults to the write time. Writes for one identity must arrive in non-decreasing validFrom order.
memoryTypeNoWhich kind of memory this is, and part of its identity. preference and profile are durable user facts, goal a desired state, instruction an operating constraint, procedure reusable steps, episode a recallable event, note general knowledge.
operationsNoA batch plan. Supply this instead of a single memory, typically from an extraction step. Every operation is validated before the first one is applied, so a malformed late operation cannot leave a half-applied batch.
sourceClassNoClaimed origin, never proof: agent_claimed (the default), tool_observed, human_confirmed, or production_verified. It weights confidence only. An unrecognised label downgrades to agent_claimed, kept verbatim in sourceRaw.

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault

No arguments

Install Server

TDQS

A4.7/5.0
Behavior5/5

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

Beyond the annotations (readOnly=false, destructive=false, idempotent=false), the description adds meaningful behavior: identical content is a NOOP, new content supersedes while keeping history, DELETE invalidates, and every call commits a revision. This clarifies the mutation model and confirms why the operation is neither idempotent nor genuinely destructive.

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?

Three dense sentences carry distinct value: what the tool does, which siblings to use instead, and core behavioral semantics. The most important action is front-loaded, and no sentence is filler.

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?

For a complex 15-parameter tool with a rich output schema and fully commented input schema, the description covers the essential selection and invocation information: single-memory vs batch plan, identity-tuple reconciliation, history preservation, and sibling routing. Remaining operational details live appropriately in the schema definitions.

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

Parameters3/5

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

The input schema is exhaustively self-documenting with 100% description coverage, so the description does not need to repeat parameter details. It adds useful framing like 'by identity tuple' and the reconcile/supersede/NOOP behavior, but does not independently clarify individual parameters beyond the schema.

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 opens with a specific verb and resource: 'Add or reconcile one scoped memory, or apply an ADD/UPDATE/DELETE/NOOP plan.' It clearly distinguishes itself from sibling tools by directing decision memories to shadowgraph_record_decision, factual observations to shadowgraph_record_fact, and reads to shadowgraph_recall.

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 explicitly states when to use sibling alternatives ('Use shadowgraph_record_decision for a choice, shadowgraph_record_fact for an observation, shadowgraph_recall to read memory back'), giving an agent clear routing signals. It also clarifies the batch-plan mode versus single-memory mode as two valid ways to invoke this tool.

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

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/LiLara-AI/shadowgraph'

If you have feedback or need assistance with the MCP directory API, please join our Discord server