VC - VibeCoder v2 is an AI-orchestrated coding agent colony designed to tackle complex software development tasks by breaking them into small, well-defined issues. It targets developers and researchers seeking efficient, high-quality AI-assisted coding, offering a novel approach that minimizes costs and maximizes agent focus through a self-improving, issue-oriented workflow.

How It Works

VC employs an AI Supervisor (Sonnet 4.5) to manage an issue workflow. This loop claims ready issues, performs AI-driven strategy and risk assessment, executes tasks via worker agents (Amp, Claude Code), analyzes results, and auto-creates follow-up issues. Quality gates (test, lint, build) are applied, with AI deciding the issue's final status. The architecture features a REPL shell, an AI Supervisor, an Issue Workflow Executor (event loop), and Worker Agents, all coordinated via an SQLite issue tracker managed by the Beads library. This design emphasizes zero framework cognition, issue-oriented orchestration, and nondeterministic idempotence for robust, interruptible workflows.

Quick Start & Requirements

• Primary install/run: Build with go build -o vc ./cmd/vc and execute ./vc.
• Prerequisites: Requires ANTHROPIC_API_KEY environment variable. Integration tests incur token costs.
• Links: Core Docs (ARCHITECTURE.md, CLAUDE.md, DOGFOODING.md), detailed design docs in docs/.

Highlighted Details

• AI Supervised Issue Workflow: Proven through 24+ dogfooding missions.
• Quality Gates: Achieved a 90.9% pass rate.
• Zero Framework Cognition: All decisions are delegated to AI, avoiding heuristics.
• Beads Integration: Delivers 100x performance improvement over previous shell-based CLIs.
• Self-Hosting: The system successfully builds and improves itself.

Maintenance & Community

The project is actively developed and self-improving through "dogfooding." No specific community links (e.g., Discord, Slack) or external contributor details are provided in the README.

Licensing & Compatibility

MIT License. No explicit restrictions are mentioned for commercial use or closed-source linking.

Limitations & Caveats

The project is in "Production (Dogfooding)" status, indicating active development and potential for ongoing changes. While the core workflow is operational and self-improving, further development is planned. The system's functionality is dependent on external AI APIs, and specific unsupported platforms or known bugs are not detailed in the README.