recall

rohitg00/agentmemory

Search past observations and learnings using hybrid BM25+vector+graph search.

What is recall?

Recall retrieves context from past sessions and observations about a topic using intelligent hybrid search. Use it when you need to find what was previously discussed, decided, or learned—triggered by phrases like "recall," "what did we do about," or "have we seen."

Search across past sessions and observations using hybrid BM25, vector, and graph search
Rank results by importance score to surface high-signal findings first
Group results by session and provenance channel (user, agent, tool, import, shared)
Return exact observation type, title, and narrative from stored memory
Suggest alternative search terms when no results match

How to install recall

npx skills add https://github.com/rohitg00/agentmemory --skill recall

Claude Code

Cursor

Windsurf

Cline

How to use recall

1.Call the skill with a search query describing what you want to recall
2.The tool returns results grouped by session, ranked by importance
3.Review high-importance results (score ≥7) first
4.If no results appear, try the suggested alternative search terms

Use cases

Good for

Find previous decisions or code patterns related to a topic (e.g., token rotation strategies)
Retrieve context from past sessions to avoid re-discussing solved problems
Locate learnings and observations across multiple projects or sessions
Disambiguate conflicting information by checking user vs. agent-inferred records
Discover what teammates have shared or documented about a subject

Who it's for

Developers working across multiple sessions or projects
Teams sharing memory and context across members
Anyone needing to avoid duplicating past work or decisions

recall FAQ

What happens if the search returns no results?

The skill suggests 2–3 alternative search terms to try. It does not guess or fabricate memories.

How are results ranked?

Results are ranked by importance score (0–10), with high-signal observations (importance ≥7) shown first.

Can I search within a specific project?

Yes, pass the `project` parameter when scoping to a specific repository.

What does the provenance channel tell me?

It indicates the source: `user` (human input), `agent` (AI inference), `tool` (from a tool), `import` (external), or `shared` (teammate's write). User records are preferred over agent inference.

How is this different from recap or session-history?

Recall searches across all sessions and observations by topic; recap and session-history show session-scoped views of the same data.

Full instructions (SKILL.md)

Source of truth, from rohitg00/agentmemory.

name: recall description: Search agentmemory for past observations, sessions, and learnings about a topic using hybrid BM25 plus vector plus graph search. Use when the user says "recall", "what did we do about", "did we ever", "have we seen", or needs context from past sessions. argument-hint: "[search query]" user-invocable: true

The user wants to recall past context about: $ARGUMENTS

Quick start

memory_smart_search { "query": "jwt refresh token rotation", "limit": 10 }

Expected output:

2 results across 2 sessions.
[importance 8] decision · "Rotate refresh tokens on every use" (session 7f3a9c21)
[importance 5] code · "limit.ts counts per-IP" (session b21d004e)

Why

Only surface what the tool returned. Never fabricate an observation, a session id, or an importance score. If nothing comes back, say so.

Workflow

Call memory_smart_search with the user's text as query and limit: 10. Pass project when the user scopes to a specific repo.
Group results by session. Records carry a provenance channel (user, agent, tool, import, shared); when results conflict, prefer user over agent inference, and flag shared records as another teammate's write.
For each observation show its type, title, and narrative.
Lead with the high-signal observations (importance >= 7).
If zero results, suggest 2-3 alternative search terms and stop. Do not guess.