Tool reference · Save & recall
memory_save_fast
Saves knowledge exactly like memory_save but always skips the quality gate and never calls an LLM.
local stdio server writes
When to use
- You are saving many records in a loop and want each call to stay fast.
- The quality gate is on in your setup, but this record must be stored regardless.
- You run without any LLM provider configured.
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
contentrequired | string | — | — |
source_format | "auto" | "conversation" | "auto" | — |
typerequired | "decision" | "fact" | "solution" | "lesson" | "convention" | — | — |
project | string | "general" | — |
tags | string[] | — | — |
context | string | — | — |
branch | string | — | — |
filter | string | — | — |
importance | "critical" | "high" | "medium" | "low" | "medium" | — |
agent_id | string | — | Optional Claude Code subagent ID (v2.1.139+) |
parent_agent_id | string | — | Optional parent agent ID (the dispatching span) |
supersede | boolean | false | Same as memory_save.supersede: retire records this one gives a new value for. |
Example
Arguments
{
"content": "Staging database runs on port 5433; production uses 5432.",
"type": "fact",
"project": "my-api",
"tags": [
"database",
"config"
]
}Result shape
{
"saved": true,
"id": 1843,
"deduplicated": false,
"mode": "fast"
}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:
same as memory_save but routes through the fast hot path (skip_quality=True, no LLM, no async-blocking). Use when you want to bypass the v10 quality gate without flipping the env flag.