| name | recall |
| description | Search persistent memory for relevant notes from past sessions — prior decisions, conventions, gotchas, and how similar work was handled. Recall is cheap and fail-open, so call it rather than guess: before starting a task, editing a file, fixing a failing test, or choosing an approach — and whenever the user refers back ("what did we decide about X", "how did we do Z"). |
Recall from persistent memory
Retrieve relevant memories before you act, whenever past work could bear on the task.
(The tool is exposed by the memory plugin's MCP server; the exact tool id is
harness-specific.)
Default to recalling. It's fast and fail-open — a miss costs nothing, a skipped
hit means repeating a mistake or contradicting an earlier decision. Recall at the
start of a task and at each decision point: opening a file, diagnosing a failing
test, picking an approach, hitting something unfamiliar — not only when asked.
- Call
recall(query, k) with a focused natural-language query. k defaults to 5;
raise it for broader context.
- Hits return ranked (best first), each with
id, content, score, tokens.
- Fold relevant hits into your reasoning; cite what you used. Empty result? Just
proceed.
Recall-only: saving happens in the background (the Daydreamer watches the session),
so there's no remember tool.