Agent Ledger is the immutable log of everything the AI fleet has decided and done. Every prompt, every tool call, every rollback. Search it like a database, replay it like a video, export it for compliance review. If an agent did it, the ledger has it.
Each ledger entry captures the inbound signal (alert, page, user request), the model and prompt used, every tool call the agent made, the LLM's reasoning, the action it took, and the verified outcome at T+5m, T+1h, and T+24h. There is no separate "explainability" feature, explainability is the ledger.
The ledger is searchable by agent, by tenant, by signal id, by tool name, by outcome, by trust delta, and by free-text query against the prompt or response. Use it during a postmortem to pull every action the agents took during the incident window. Export to JSON or CSV for compliance.
Replay mode lets you scrub through a single decision the way a flight recorder plays back a flight. See the inbound signal, watch the agent's thinking step by step, see each tool call resolve, and see the verifier mark the outcome. Use it to teach a new on-call engineer or to debug a regressed agent.
The ledger is built for audit. Each entry is hashed and chained, so altering a past entry breaks the hash chain and is detectable. Exports include the chain proof. Retention is configurable per tenant (default seven years) and the schema is documented for your auditor up front.
Subscribe to Nova AI Ops on YouTube for demos, tutorials, and feature deep-dives.
When an auditor asks why the agent did something at 3am, you have the receipt. Prompt, model, tool calls, outcome, rollback if any.