Use Kernall
Audit logs
Inspect original policy verdicts and their linked human reviews.
Read action history
# Local development: read the latest checks awaiting a human verdict.
curl --fail-with-body --get "$KERNALL_BASE_URL/api/logs" \
--data-urlencode 'decision=requires_human' \
--data-urlencode 'limit=50'The response contains logs and next_before. Records are newest first and include ID, timestamp, project ID, action, destination, context, detected category, verdict, reason, and optional confirmation ID.
Filters and pagination
| Parameter | Behavior |
|---|---|
| decision | approved, denied, or requires_human. Other values return 400. |
| from / to | Inclusive bounds; ISO 8601 or positive Unix milliseconds. |
| before | Return timestamps strictly earlier than the cursor. |
| limit | Default 50; bounded to 1–200. |
| next_before | Response cursor when a full page is returned; otherwise null. |
Read more records carefully
Pass next_before as the next request’s before, keeping other filters unchanged. A full last page may still return a cursor; its following page can be empty.
The cursor uses only a timestamp. If more records share a boundary millisecond than fit on the page, later pages can skip tied records. This pagination is not a guaranteed complete audit export. Dedicated exports are not implemented.
Use valid dates. Invalid timestamp parameters are currently ignored.
Interpret the record
API audit verdicts are immutable. Resolution is recorded separately; follow confirmation_id to read its final status. An approved proposal is not evidence that your agent completed the operation.
Raw data is absent, but destination and context are stored. Keep secrets out of those fields. See data handling.