Tool reference · Errors, rules & self-improvement
self_patterns
Summarises logged errors, insights and rules: repeating error categories, insights ready for promotion, rule success rates and the weekly error trend.
local stdio server read-only
When to use
- You want to know which kinds of mistakes keep happening.
- You want to find insights that are ready to be promoted to rules.
- You want to check whether errors are going down week over week.
Parameters
| Name | Type | Default | Description |
|---|---|---|---|
view | "error_patterns" | "insight_candidates" | "rule_effectiveness" | "improvement_trend" | "full_report" | "full_report" | — |
project | string | — | — |
days | integer | 30 | — |
Example
Arguments
{
"view": "improvement_trend",
"project": "my-api"
} Result shape
{
"improvement_trend": {
"weekly_errors": [
{
"week_ago": 0,
"errors": 2
},
{
"week_ago": 1,
"errors": 4
},
{
"week_ago": 2,
"errors": 3
},
{
"week_ago": 3,
"errors": 6
}
],
"direction": "improving"
}
} full_report returns error_patterns, promotion_candidates, rule_effectiveness, improvement_trend and summary together.
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:
Analyze error patterns and self-improvement stats. Views: error_patterns (frequency, repeating 3+), insight_candidates (ready for promotion), rule_effectiveness (success rates, stale rules), improvement_trend (weekly errors), full_report (all). Call periodically to track improvement.