Micro Agent is an open-source CLI tool by Builder.io that generates and iterates on code using LLMs until all specified unit tests pass. Unlike general-purpose coding assistants, it is narrowly scoped: given a description and test file, it writes code and re-runs the tests in a loop until every test passes. It supports JavaScript, TypeScript, and Python, and works with Claude, OpenAI, Ollama, or any OpenAI-compatible provider. It also integrates with Figma for design-to-code conversion with visual accuracy checks. The project lives on GitHub and is free to use. Key features: - Test-driven code generation: iterates until all unit tests pass, not just until output looks right - Supports JavaScript, TypeScript, and Python - Compatible with Claude, OpenAI, Ollama, and any OpenAI-compatible LLM provider - Figma integration for design-to-code with visual matching verification - Open-source and free (bring your own LLM API key) - Focused single-purpose scope reduces compounding errors common in general agents
Free and open-source (MIT license). You pay only for LLM API calls using your own API key.