Tool reference · Tasks, workflows & decisions
New in 14.9.0memory_exploration_recall
Finds earlier exploration conclusions for a codebase question and checks each source file against its recorded hash, so you know whether a conclusion can be reused as is.
local stdio server read-only
When to use
- You are about to read files to answer a codebase question and want to check whether it was answered before.
- You want to know which files changed since an earlier conclusion was recorded, to re-read only those.
- You want only conclusions that are still valid (include_stale: false).
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
questionrequired | string | — | The question being asked now (min length 1, max length 2,000) |
projectrequired | string | — | Project name (min length 1, max length 200) |
root | string | — | Absolute project root to verify files against; default: the server's working directory (max length 1,024) |
limit | integer | 5 | Maximum memos returned (min 1, max 50) |
include_stale | boolean | true | Also return memos whose files changed since recording |
Example
Arguments
{
"question": "where is auth handled?",
"project": "my-api",
"root": "/Users/dev/projects/my-api",
"limit": 5
} Result shape
{
"question": "where is auth handled?",
"project": "my-api",
"root": "/Users/dev/projects/my-api",
"fresh": 0,
"stale": 1,
"results": [
{
"id": 12,
"question": "how is auth wired?",
"conclusion": "JWT validation runs in middleware/auth.py before every /api route; tokens are issued in services/tokens.py and refresh tokens live in the refresh_tokens table.",
"status": "stale",
"files": [
"middleware/auth.py",
"services/tokens.py"
],
"recorded_at": "2026-10-13T09:41:07.512000Z",
"updated_at": "2026-10-13T09:41:07.512000Z",
"changed_files": [
"services/tokens.py"
],
"missing_files": [],
"score": 0.71
}
]
} Results are ranked by question similarity. A fresh memo has every source file byte-identical to when it was recorded; a stale one lists changed_files and missing_files to re-read.
Values are illustrative; the keys follow the server's handler. MCP clients receive the result as JSON text content.
Server description
The description the server sends to your agent in tools/list, captured from the v14.9.0 source:
Before re-reading files to answer a codebase question, look up earlier exploration conclusions. Results are ranked by question similarity; each carries status 'fresh' (every source file is byte-identical to when the conclusion was recorded: reuse it) or 'stale' (changed_files / missing_files list what to re-read).