List recent agent sessions with aggregated stats — call count, total cost, token usage, and start time. Use this to find a session_id before calling get_session or to get a cost overview.

Retrieve the full captured trace for a single LLM call. Returns the prompt, system prompt, tool calls, model response, token usage, cost, and latency.

Retrieve the LLM calls of an agent session in chronological order. By default each call is a compact summary (index, model, status, error reason, tokens, cost, latency, tool names, sizes) — sessions can be megabytes, so full prompt/response bodies are returned only with include_messages=true, and single calls are better fetched via the explain tool using the action_id from a summary row.

Full-text search across all captured LLM calls. Searches prompts, outputs, system prompts, agent names, and user IDs. Use this to find calls related to a topic, error, or agent.

List loop runs (explicit x-agenticledger-run-id or auto-inferred fresh-context loops, e.g. Ralph overnight runs) with iterations, sessions, cost, flagged-call counts, and status (running / flagged / complete).

Status of one loop run: iterations so far, total cost and tokens, flagged calls, and whether the completion promise was seen (status=complete). Loop runners and agents can use this to decide whether to continue iterating.