mabbux
v0.0.2
Published
A governable harness for AI coding agents: code-anchored memory, verified evidence and CI-enforced rules.
Maintainers
Readme
mabbux
A governable harness for AI coding agents.
⚠️ Early development. Not ready for use yet.
mabbux gives the coding agent you already use (Claude Code, GitHub Copilot, Cursor, Codex…) what the model alone lacks, and makes it verifiable:
- Code-anchored memory. Project knowledge in open formats (AGENTS.md, OKF), linked to the code it describes and flagged as stale when that code changes.
- Evidence for every task. A task is done when its proof has been run and recorded, not when the agent says so.
- Rules enforced outside the model. The same checks run in agent hooks, git hooks and CI, so they hold for any assistant.
- Built on open standards. Agent Skills, MCP, OKF and OpenTelemetry. No lock-in to a single assistant.
Designed for solo developers first, with team and organization features (shared policies, signed skills, audit trails) to follow.
Status
The design is in progress. Follow the repository for the first release.
License
MIT
