Your agents' flight recorder. Proof of what your AI did.
- Prove the log wasn't tampered with. Every action is hash-chained the moment it's recorded. One click verifies the whole ledger — and any edit breaks the chain at the exact link.
- Catch rogue behavior live. Burst spikes, first-seen tools, and after-hours activity trigger alerts the moment they happen.
- Hand auditors evidence, not debug dumps. Printable audit reports, signed Tamper-Seal certificates, and a public integrity badge for your trust page.
Free for $0/mo · no credit card · one-line SDK
Agent Flight Recorder is a tamper-evident audit trail for AI agents: every action, tool call and decision is hash-chained with SHA-256 into an append-only ledger, so you can prove what your AI did after an incident. Built by Ingenious AI.
Every link commits to the last.
Rewrite one record, break them all.
- Your agent posts every action — tool calls, decisions, outputs — to
/v1/actionswith an API key. One line of Python or JavaScript, any framework. - Each record is hash-chained — record N stores SHA-256 of record N-1's hash plus its own payload. There is no edit button; there is only the chain.
- Verify in one click — the console recomputes every hash server-side and stamps PASS/FAIL with the exact break point. Evidence you can hand to an auditor.
- Prove it publicly — embed the integrity badge on your trust page: green chain intact · N actions, or red the moment anything breaks.
Langfuse and LangSmith debug traces.
We produce proof.
Agent observability answers "what is my agent doing?" for developers. Agent Flight Recorder answers the question incidents, enterprise buyers, and regulators actually ask: "can you prove what your agent did?"
That needs a different instrument: an independent, append-only, hash-chained ledger with one-click verification, a public integrity badge, anomaly alerts, and SOC2-shaped audit reports — not prettier traces.
Tamper-evident by construction
Append-only hash chain. No developer, admin, or attacker can rewrite history silently.
Public integrity badge
A live SVG on your trust page — the world sees your trail is intact.
Audit-grade reports
Scope, methodology, findings, chain integrity statement. Print-ready.
The recorder watches while it records.
Every action is screened by rule-based anomaly detection the moment it lands:
Burst > 3σ
Action volume beyond three standard deviations of the trailing 24-hour baseline fires an alert.
First-seen tool
Your agent touches a tool it has never used before? That's an inbox alert, instantly.
After-hours spike
Activity at 00:00–05:00 UTC against a 30-day baseline gets flagged.
Incident replay
Step through the action sequence with timestamps and reconstruct exactly what happened, in order.
Evidence instruments nobody else has.
These exist because of the hash-chained ledger. Observability tools trace mutable debug data — there is nothing tamper-evident to build on, so they can't offer any of this:
One click mints a signed, printable certificate of ledger integrity — agent, action count, chain tip, HMAC-SHA256 signature. Auditors verify it without an account, on a public page or JSON API. We never certify a broken chain.
Agents declare caused_by links at ingest; we validate and pin the exact
record. The graph renders every causal path across your agents as a labeled DAG — trace a
multi-agent incident back to its cause with hash citations.
Interrogate the ledger in plain language — "what happened before seq 42", "how many tool_calls". Computed deterministically, never by an LLM, so it can't hallucinate; every claim carries a seq # and hash.
Proof is free to start.
Pro
$49/mo
10 agents · 100k actions/mo · CSV export · audit reports · anomaly alerts · replay
Questions, answered.
What is Agent Flight Recorder?
Agent Flight Recorder is a tamper-evident audit trail for AI agents. Your agent posts every action — tool calls, decisions, outputs — to our API, and each record is hash-chained to the previous one. The result is an append-only ledger that proves what your AI did, which no one can silently edit afterwards.
How does the hash chain prove the ledger was not tampered with?
Each action record stores the SHA-256 hash of the previous record's hash plus its own payload. Change a single character in any old record and its hash changes, which breaks every later link. The dashboard's Verify chain button recomputes every hash server-side and reports PASS/FAIL with the exact break point.
How is this different from Langfuse or LangSmith?
Langfuse and LangSmith are debugging observability tools built to help developers inspect traces. Agent Flight Recorder is built for proof: an independent, append-only, hash-chained ledger whose job is post-incident evidence. It adds one-click chain verification, an embeddable integrity badge for trust pages, and SOC2-shaped audit reports.
What can I do with the free plan?
The free plan covers 1 agent and 1,000 actions per month, including the full hash-chained ledger, one-click chain verification, and the embeddable integrity badge. CSV export, printable audit reports, anomaly alerts, and incident replay are on Pro ($49/month) and Team ($149/month).
How do I integrate it with my agent?
Create a workspace and an API key, then POST each action to /v1/actions with a Bearer key — agent_id, action_type, tool_name, input/output summaries, and metadata. One line of Python or JavaScript does it; see the docs page for copy-paste snippets that work with any agent framework.
What are Tamper-Seal certificates?
Tamper-Seal certificates are cryptographic attestations of ledger integrity, available on Pro and Team. One click certifies an agent's ledger — action count, chain tip, minted timestamp — signed with HMAC-SHA256. Anyone can verify the certificate without an account on a public page or JSON API; a broken chain can never be certified. They exist because the ledger is hash-chained: no observability tool with mutable traces can offer this.
What happens when something looks wrong?
Rule-based anomaly detection runs on every recorded action: action bursts beyond 3 standard deviations, first-seen tool usage, and after-hours activity spikes all raise in-app alerts. The incident replay view then lets you step through the agent's action sequence with timestamps to reconstruct exactly what happened.
The next incident is a matter of when.
Be the team with the tape.
Start your AI agent audit trail in one API call. Free forever plan — no credit card.
Start recording freeFlight Recorder Pro — $49/mo: 10 agents, CSV export, audit reports, anomaly alerts.