Tool reference · Sessions & episodes
memory_episode_save
Saves a short narrative of how a piece of work went: what was tried, what failed, what worked, and the key insight.
local stdio server writes
When to use
- You just solved a hard problem and want to remember the path, not only the answer.
- An approach failed and you want future sessions to know why.
- You want to tag the experience with concepts so it can be found by topic later.
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
narrativerequired | string | — | 2-3 sentence narrative of what happened |
outcomerequired | "breakthrough" | "failure" | "routine" | "discovery" | — | — |
project | string | "general" | — |
impact_score | number | 0.5 | 0.0-1.0, how significant |
concepts | string[] | — | Key concepts involved |
approaches_tried | string[] | — | — |
key_insight | string | — | The aha moment, if any |
frustration_signals | integer | 0 | — |
Example
Arguments
{
"narrative": "Webhook retries were duplicating orders. Tried a DB unique constraint first, then added an idempotency key header check, which fixed it.",
"outcome": "breakthrough",
"project": "my-api",
"impact_score": 0.8,
"concepts": [
"webhooks",
"idempotency"
],
"approaches_tried": [
"unique constraint on order_ref",
"Idempotency-Key header"
],
"key_insight": "Deduplicate at the request boundary, not in the orders table."
}Result shape
{
"saved": true,
"episode_id": "a3f9c0d1e2b34f5a9c8d7e6f5a4b3c2d"
}When concepts are given, the narrative is also run through the concept extractor and linked into the knowledge graph.
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:
Save an episode — narrative of WHAT HAPPENED and HOW. Not just facts, but the journey: what was tried, what failed, what worked.