recall

name	recall
description	Search mempalace for past observations, sessions, and learnings about a topic
argument-hint	[search query]
user-invocable	true

The user wants to recall past context about: $ARGUMENTS

Quick start

Call memory_smart_search with the user's query (limit 10).

Why

Retrieve context from past sessions so the agent can answer accurately without hallucinating.

Workflow

1. Call memory_smart_search with query from $ARGUMENTS and limit: 10.
2. Group results by session ID.
3. For each observation, show type, title, and narrative.
4. Highlight observations with importance >= 7.
5. If empty, suggest 2-3 alternative search terms.

Anti-patterns

WRONG -- making up observations to fill gaps:

// No results from tool, but agent fabricates "I recall we discussed X"

RIGHT -- honest empty state with suggestions:

// No results found. Try: "deployment config", "build system", "cli flags"

WRONG -- ignoring the limit parameter:

// Calling memory_smart_search without limit, getting hundreds of results

RIGHT -- using explicit limit:

memory_smart_search({"query": "config file parsing", "limit": 10})

MCP tool unavailable

If memory_smart_search is not available:

1. Run /plugin list in Claude Code, confirm mempalace is enabled.
2. Restart Claude Code (.mcp.json is only read on startup).
3. Check /mcp to see if the mempalace MCP server is connected.

See _shared/TROUBLESHOOTING.md and EXAMPLES.md for more.