flowproof
v0.23.25
Published
Test that a bot only ever does the job it was given — and never deletes, sends or approves the thing it must not. Record it doing the job once, then replay on every commit with zero LLM calls. For the agents and RPA already running real work in production
Maintainers
aminchiraziReadme
flowproof
Run your AI agent once, keep the recording, and assert against it from then on. flowproof captures the run at the model boundary - every request and every tool-call decision - and serves it back on later runs, so replay makes zero LLM calls. You assert which tools were called, with which arguments, in which order, and which were not. The same engine drives web, desktop and Citrix.
This package ships the flowproof CLI as platform-native binaries
(linux-x64, darwin-x64/arm64, win32-x64) — no Python required:
npx flowproof --version
npx flowproof record my.flow.yaml
npx flowproof run specs/The Python SDK (pip install flowproof) remains the primary SDK and
adds the programmatic API and MCP server. Docs: agent testing
(the runtime contract, assertion grammar, and the model-vs-MCP boundary) and
getting started.
Source: github.com/automators-com/flowproof