Retrace
Verifiedby Retrace
Records, replays, forks, and shares AI agent runs — forks re-issue only LLM calls while tool calls replay from tape, turning production failures into regression tests and PR eval gates.
- Access
- CLIAPI
- Auth
- API Key
- Skills
- 3
Skills
Record & Replay
Records agent runs and replays them deterministically, re-issuing only LLM calls while tools play from tape.
Fork Runs
Forks a captured run to test prompt or code changes against the exact same production context.
Eval Gates
Turns captured failures into regression tests and PR eval gates that block regressions in CI.
Related Agents
promptfoo
by promptfoo
Tests and red-teams LLM apps, agents and RAG pipelines from declarative config, scanning for prompt injection…
Argent
by Software Mansion
Gives AI coding assistants control of iOS simulators, Android emulators, TVs, and Electron apps to tap, debug, profile…
MCPJam
by MCPJam
Tests and evaluates MCP servers and MCP apps for developers, with a playground, OAuth debugger, cross-client evals, and…
DeepEval
by Confident AI
Unit-tests LLM and agent outputs in a pytest-style workflow, scoring answer relevancy, faithfulness, hallucination and…