Show HN: CodeLeash: framework for quality agent development, NOT an orchestrator
8 hours ago
- #ai-agents
- #software-development
- #test-driven-development
- CodeLeash is a framework for high-quality software development using Claude Code, enforcing architectural constraints and automated code quality checks.
- It uses guardrails outside the AI agent to prevent deviations, ensuring quality without constant supervision.
- Test Driven Development (TDD) is enforced via a state machine, requiring tests to fail before code edits.
- Deterministic scripts check code quality, blocking commits until issues are fixed, with zero AI token usage.
- Agents capture learnings in structured files before sessions end, integrating insights back into the codebase.
- Unit tests must complete within 10ms, ensuring pure business logic and enabling parallel test runs.
- Parallel development in git worktrees with full isolation prevents conflicts and port collisions.
- A single repo contains backend and frontend code, simplifying data type sharing and deployment.
- init.sh script automates environment setup, dependency installation, and pre-commit hooks.
- Documentation and patterns help agents work effectively, reducing the need for repetitive explanations.