What it does
Refact.ai is an open-source AI coding agent that runs in the IDE. You describe a task and the agent plans and carries it out step by step, searching the repository and connecting to tools such as GitHub, databases, Docker and CI/CD pipelines, while you can preview and control the process. Alongside the agent there is in-IDE chat for asking, editing and debugging code, and real-time autocompletion that the site says is powered by Qwen2.5-Coder with retrieval-augmented generation. The site lists plugins for VS Code, JetBrains IDEs and Visual Studio, with Neovim and Sublime Text marked beta, and says it supports more than 25 programming languages. Chat and the agent can use models such as Claude 4, GPT-4o and GPT-4o mini, or you can bring your own API key for providers like Gemini, Grok, OpenAI and Deepseek. A self-hosted, on-premise version keeps code on your own infrastructure, and an enterprise offer adds fine-tuning on a company's codebase and priority support. A banner at the top of the site says Refact Cloud is shutting down soon and links to a blog post about it. The home page does not explain how the notice affects the plans it lists. A few free online tools for code generation, code review and image-to-code are also linked.
Capabilities
- Autonomous agent that plans and executes coding tasks inside the IDE
- In-IDE chat plus autocompletion powered by Qwen2.5-Coder and RAG
- Connects to GitHub, databases, Docker and CI/CD pipelines
- Choose hosted models or bring your own API key
- Self-hosted and on-premise deployment
- Plugins for VS Code, JetBrains IDEs, Visual Studio, Neovim and Sublime Text
When to use it
- Handing a coding task to an agent that works in your own IDE
- Keeping code on your own servers with a self-hosted setup
- Using your own LLM API keys instead of a fixed model
