Skip to main content

Agent actions

GET 
/api/ledger/agent-actions

Agent Audit Ledger: one row per GOVERNED AGENT ACTION, not per DB write.

The append-only chain (list_ledger above) records portal state changes; this is the operational view a reviewer actually audits — what each agent tried to do, against which system, how policy decided, who approved it and what happened — assembled from the real enforcement record:

gateway_requests — every LLM action the AI Gateway saw (decision, policy, model, tokens, latency, redacted prompt/response) gateway_db_queries — the SQL an agent ran during that work approval_requests — who approved/denied a held action audit_ledger — the hash-chain entry covering the same event

Nothing is synthesised: a field with no source is returned null so the UI can show "—" rather than invent a value.

Request​

Responses​

Successful Response