Tool reference · Save & recall
memory_get
Fetches the full content of specific records by ID, up to 50 at a time.
local stdio server read-only
When to use
- You ran memory_recall with mode: 'index' and now want the full text of a few chosen hits.
- Another tool returned record IDs (history, relations, evidence) and you need their content.
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
idsrequired | integer[] | — | Knowledge record IDs (max 50 per call; extras are silently dropped) |
detail | "summary" | "full" | "full" | 'summary' truncates content to 150 chars, 'full' returns everything |
Example
Arguments
{
"ids": [
1842,
1843
],
"detail": "full"
}Result shape
{
"total": 1,
"detail": "full",
"results": [
{
"id": 1842,
"type": "decision",
"content": "Use PostgreSQL advisory locks for the nightly billing job instead of a Redis lock.",
"project": "my-api",
"tags": [
"database",
"billing"
],
"created_at": "2026-09-20T10:14:03Z",
"context": "Redis lock expired mid-run twice…",
"session_id": "a1b2c3",
"status": "active",
"confidence": 1,
"last_confirmed": "2026-09-24T08:01:55Z",
"recall_count": 3,
"branch": "main"
}
]
}Unknown IDs are skipped silently, and IDs beyond the first 50 are dropped.
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 v14.7.0 source:
Batched fetch by ID — complement to memory_recall(mode='index'). Returns full content for ONLY the IDs the caller chose after inspecting an index. Typical 3-layer flow: recall(mode='index') → pick IDs → memory_get(ids=[...]).