Agent Flight Recorder

Cross-provider incident
repair for AI agents.

When a production run fails or burns cost, Lieva records the trace, pinpoints the broken step, replays the fix, and turns the incident into a regression eval.

Forks run locally. Your keys never leave your env.

Coming Soon

Early access opening

Early access

Founder seats only.
The first 50 teams.

We are building the forensic layer for the agentic era. Founding members get direct engineering support and shape the replay-engine roadmap.

Founder seats are free to claim while they last →

Founding member · first 50

Founder access

Freewhile seats last

No charge before launch — plus 40% off the Founder plan, locked for life.

  • Guaranteed early access the moment the preview opens
  • A live 1:1 session — we repair one of your real failed runs with you
  • 40% off the Founder plan — locked for life
  • Private founder channel + roadmap influence

No card required

FAQ

Answers before you commit.

No. Forks and replays run locally in your own environment — your provider keys never leave your machine. Lieva works on the trace you import.

TypeScript SDK first, Python next: OpenAI, Anthropic, the Vercel AI SDK, Mastra, and Trigger.dev / Inngest.

Observability shows you the trace. Lieva turns a bad run into a fixed replay and a regression eval — it is the post-incident repair loop, not another dashboard.

Guaranteed early access, a live 1:1 session where we repair one of your real failed runs with you, 40% off the Founder plan locked for life, and a private founder channel with roadmap influence. Claiming a seat is free — just join the waitlist.