traceforge

English | 简体中文

> Open-source replay and regression testing for AI agents. > > Capture real runs. Replay failures. Export tests.

Traceforge makes agent runs reproducible.

It records what actually happened across terminal, MCP, and browser tools, packages the run into a portable trace bundle, helps you inspect where it failed, and turns real failures into regression fixtures and tests.

Modern agent workflows are powerful, but hard to trust.

• terminal output is incomplete
• MCP calls are buried in protocol logs
• browser failures are hard to reproduce
• expensive incidents rarely become tests
• teams fix failures once, then see them return later

Traceforge solves one narrow problem well:

capture -> inspect -> replay -> export test

It is not another agent framework. It is the verification layer for the agent ecosystem.

• unified traces from terminal, MCP, and browser runs
• a local trace viewer with timeline, artifacts, and failure boundaries
• replay that re-runs a flow and checks where it diverges
• export to reusable regression fixtures and test templates
• shareable trace bundles for debugging and collaboration

traceforge capture -- node -v
traceforge view ./.traceforge/traces/run_xxxxxxxx
traceforge replay ./.traceforge/traces/run_xxxxxxxx
traceforge export-test ./.traceforge/traces/run_xxxxxxxx

Traceforge is currently in early alpha.

Requirements:

• Node.js 20+
• pnpm 10.6.0+

pnpm install
pnpm build

Run the main verification commands:

pnpm typecheck
pnpm test
pnpm test:coverage
pnpm build
pnpm test:generated

Useful local entry points:

pnpm dev:cli
pnpm dev:viewer

• apps/cli - CLI entrypoint and local trace bridge
• apps/viewer - local viewer UI
• packages/schema - canonical trace schema and capability contract
• packages/core - bundle, replay, and v