A flight recorder for agents
Live thought streams and run timelines — watch any run, audit any run.

Without memory
When an autonomous agent does something odd, the evidence is usually gone — stdout scrolled away, no record of what it was thinking, no timeline of what happened when. Debugging becomes guessing, and trust in automation erodes.
With Rekall
Every run writes a permanent flight record: thoughts, steps, checkpoints, and status changes, streamed live over SSE and stored for later. Watch a run as it happens or reconstruct one from last month, step by step.
How it works
Runs narrate themselves
Agents record thoughts and progress as they work; the run-event bus publishes every state change.
Watch live
The dashboard tails active runs over SSE — no polling, no lost frames, streams survive split API/worker deployments.
Audit later
The full timeline persists: what the agent believed, what it did, what it cost, and where it paused for approval.
See it in action
curl -N https://api.rekall.ai/api/v1/agent-constellations/\$AGENT/runs/$RUN/events \-H "Authorization: Bearer rk_..."data: {"type":"progress","thought":"batch 3/9 done"}data: {"type":"progress","thought":"row 812k, on pace"}data: {"type":"status","status":"paused","reason":"gate: DROP TABLE requires approval"}
Related use cases
Durable agent runs — checkpoint, resume, audit
A crashed agent resumes from its last checkpoint, not from scratch.
Human-in-the-loop approval gates
Autonomous until it matters. Then a human decides — and the agent resumes.
Agent-to-agent collaboration over A2A
Your LangGraph pipeline delegates to your CrewAI crew — over an open protocol.