An AI coding agent that plans, builds, and verifies — your choice of model.

Claude architects the work. Qwen implements it, one task at a time. You approve every change. Self-hosted, provider-agnostic (Qwen · Claude · OpenAI).

Architect

Claude

Reads your codebase and produces a validated Implementation Playbook — tasks, acceptance criteria, risks.

Implementer

Qwen

Executes the playbook task by task: proposes diffs, applies them, and runs the build to verify.

Owner

You

Pick the provider, review each diff, and gate every task. The agent works on your server, on your code.

mgc run --mode plan --provider anthropic "Add CSV export to the reports API"
mgc playbook validate playbook.md      # checked against a frozen schema
mgc playbook run     playbook.md       # Qwen implements, you approve, build verifies

Layered .NET 8 backend · PostgreSQL · CLI-driven · propose / apply / VERIFY with auto-fix.