mini-code-agent-langgraph 0.3.2
0.3.2 tightens the trust boundary around verified patches and documents a CI-verified release path.
Highlights
- Recognized zero-test runs fail closed by default;
--allow-zero-testsis an explicit weakening flag. mca runnow requires an explicit model and authoritative test command, while chat without a test command stays in read-only/askmode.- The deterministic offline policy suite covers 11 cases: nine verified submissions and two expected refusals, with zero unexpected submissions and zero unrelated changes.
- GitHub Actions, dependency review, CodeQL, pinned actions, dual dependency audits, protected
main, and protected release tags harden the supply chain.
The 11/11 result measures offline runtime-policy conformance with scripted local decisions. It does not measure model quality, autonomous repair ability, provider behavior, real-world task success, or SWE-bench performance.
Install and try without an API key
python -m pip install mini-code-agent-langgraph==0.3.2
mca demoRun mca demo on macOS, Linux, or WSL2. Native Windows supports informational CLI and configuration paths only.