Tool reference · Errors, rules & self-improvement
self_rules
Manages behavioural rules: list them, record when one was used, rate whether it helped, and suspend, reactivate, retire or add rules by hand.
local stdio server writes
Caution. A rule rated after 10 or more uses with a success rate below 0.2 is suspended automatically. retire and suspend remove a rule from self_rules_context results.
When to use
- You want to see the active rules for a project.
- After a task, you want to rate whether a rule actually helped.
- You want to add a rule directly without going through insights (action "add_manual").
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
actionrequired | "list" | "fire" | "rate" | "suspend" | "activate" | "retire" | "add_manual" | — | — |
id | integer | — | Rule ID (for fire/rate/suspend/activate/retire) |
success | boolean | — | For rate: was rule helpful? |
content | string | — | Rule text (for add_manual) |
category | string | — | Category (for add_manual) |
scope | string | "global" | global | project:<name> | category:<name> |
priority | integer | 5 | 1-10 |
project | string | "general" | — |
tags | string[] | — | — |
Example
Arguments
{
"action": "add_manual",
"content": "Always run the migration in a transaction on staging first.",
"category": "config_error",
"scope": "project:my-api",
"priority": 7,
"project": "my-api"
}Result shape
{
"added": true,
"id": 19
}Other actions return their own small objects, for example rate returns {rated, id, success_rate}.
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:
Manage behavioral rules (SOUL). Rules are promoted insights that shape agent behavior. Actions: list, fire (record relevance), rate (success=true/false), suspend, activate, retire, add_manual. Auto-suspend: success_rate < 0.2 after 10+ fires.