Skip to main content
Glama
sleepunit-agents

loom

by sleepunit-agents
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Local

remember

Store an episodic memory that persists across sessions. Use it to save important facts about the user, a project, or yourself so they remain available in future conversations.

Instructions

Store an episodic memory that persists across sessions. Use this when you learn something important about the user, a project, or yourself that should be available in future sessions.

Input Schema

NameRequiredDescriptionDefault
ttlNoTime-to-live: "7d", "30d", "24h", "permanent", or omit for no expiration.
scopeNoOverride this memory's recall visibility. Omit to take the category default: "user" and "reference" default global, everything else defaults to project-scoped when a project is set (global when it isn't — nothing to scope against). Set this explicitly to "global" for a fact learned in one project that actually applies everywhere — a security, git, or tool-workflow practice, not a language/framework/file-layout/code-style/error-handling convention (those stay project-scoped).
titleYesShort title for the memory
contentYesThe memory content — what you learned, observed, or were told
projectNoAssociated project, if any (omit for global memories)
categoryYesMemory category: user (about the human), project (about work), self (capability/learning), feedback (corrections/confirmations), reference (external pointers), pursuit (active goal or ongoing creative thread), episode (short-term cross-body tape: where you were / what was said or decided / what shipped / what is open — 48h TTL by default, set metadata.where to your surface e.g. "discord:#general", "voice", "wake:<id>", "lane:tending", "terminal")
metadataNoArbitrary key-value metadata
sourcingNoProvenance/trust tier — how this memory was produced. "observed": directly witnessed (measured, tested, read from source). "relayed": learned from another person/body (conversation, letter, episode). "inferred": AI-generated reasoning/synthesis — treat with skepticism, not ground truth. "system": auto-written by the harness (episode fallback, lane output). Shown in recall results and the boot digest so readers calibrate confidence.
provenanceNoFree-form origin: surface, body, session, or person. E.g. "wake:w-74fccc 2026-09-07", "Jonathan, Discord #loom", "lane:letters". For episodes, prefer metadata.where (already conventional); provenance carries detail.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed4 schema fields changedv0.5.1
    • changedInput schema / $schema
      Previous value: -"http://json-schema.org/draft-07/schema#"New value: +"https://json-schema.org/draft/2020-12/schema"
    • addedInput schema / properties / provenance
      Added value: +{
+  "description": "Free-form origin: surface, body, session, or person. E.g. \"wake:w-74fccc 2026-09-07\", \"Jonathan, Discord #loom\", \"lane:letters\". For episodes, prefer metadata.where (already conventional); provenance carries detail.",
+  "type": "string"
+}
    • addedInput schema / properties / scope
      Added value: +{
+  "description": "Override this memory's recall visibility. Omit to take the category default: \"user\" and \"reference\" default global, everything else defaults to project-scoped when a project is set (global when it isn't — nothing to scope against). Set this explicitly to \"global\" for a fact learned in one project that actually applies everywhere — a security, git, or tool-workflow practice, not a language/framework/file-layout/code-style/error-handling convention (those stay project-scoped).",
+  "enum": [
+    "project",
+    "global"
+  ],
+  "type": "string"
+}
    • addedInput schema / properties / sourcing
      Added value: +{
+  "description": "Provenance/trust tier — how this memory was produced. \"observed\": directly witnessed (measured, tested, read from source). \"relayed\": learned from another person/body (conversation, letter, episode). \"inferred\": AI-generated reasoning/synthesis — treat with skepticism, not ground truth. \"system\": auto-written by the harness (episode fallback, lane output). Shown in recall results and the boot digest so readers calibrate confidence.",
+  "enum": [
+    "observed",
+    "relayed",
+    "inferred",
+    "system"
+  ],
+  "type": "string"
+}
  2. Changed2 schema fields changedv0.5.0
    • changedInput schema / properties / category / description
      Previous value: -"Memory category: user (about the human), project (about work), self (capability/learning), feedback (corrections/confirmations), reference (external pointers), pursuit (active goal or ongoing creative thread)"New value: +"Memory category: user (about the human), project (about work), self (capability/learning), feedback (corrections/confirmations), reference (external pointers), pursuit (active goal or ongoing creative thread), episode (short-term cross-body tape: where you were / what was said or decided / what shipped / what is open — 48h TTL by default, set metadata.where to your surface e.g. \"discord:#general\", \"voice\", \"wake:<id>\", \"lane:tending\", \"terminal\")"
    • changedInput schema / properties / category / enum
      Previous value: -[
-  "user",
-  "project",
-  "self",
-  "feedback",
-  "reference",
-  "pursuit"
-]New value: +[
+  "user",
+  "project",
+  "self",
+  "feedback",
+  "reference",
+  "pursuit",
+  "episode"
+]
  3. First observedv0.4.1

TDQS

A3.7/5.0
Behavior3/5

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

Annotations already declare write (readOnlyHint=false) but non-destructive, non-open-world behavior, so the safety profile is covered. The description adds one genuine behavioral fact — persistence across sessions — but says nothing about deduplication, whether the write is immediately recallable, or how it interacts with the propose/ratify flow.

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?

Two sentences, both load-bearing: the first defines the operation and its durability, the second routes the agent to the moment of use. Nothing is repeated from the schema or annotations.

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

Completeness3/5

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

For a nine-parameter tool with nested metadata, three enums and no output schema, the description covers purpose and timing but not the surrounding workflow — no note on duplicate handling, project vs global defaults at a high level, or what a subsequent recall will show. The schema fills most of the gap, so this is adequate rather than complete.

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?

Schema description coverage is 100% and the parameter docs are unusually rich (ttl formats, scope defaults, category semantics, sourcing trust tiers), so the description is not required to carry parameter meaning. It contributes no parameter guidance of its own, matching the baseline for fully documented schemas.

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

Purpose4/5

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

States a specific verb and resource ('Store an episodic memory') plus the defining property of persistence across sessions, so the agent knows this is a durable write, not a read or ephemeral note. It does not differentiate itself from adjacent writers like memory_propose or knowledge_write, and calling a seven-category memory tool 'episodic' slightly undersells its scope.

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

Usage Guidelines4/5

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

Gives an explicit trigger condition ('when you learn something important about the user, a project, or yourself that should be available in future sessions'), which is enough to recognize the right moment. It offers no exclusions or alternatives, so the agent gets no help choosing between this and memory_propose/knowledge_write.

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

Deploy Server

Other Tools