Built for developers who want an agent they can trust.
What CodeBuddy is
CodeBuddy is a coding agent that lives in your editor — VS Code, Cursor, Windsurf, and VSCodium. It reads your codebase, plans a change, edits files across your project, and runs terminal commands — real multi-step work, not just single-line autocomplete. It's free and open source, and it works with nine LLM providers, including fully local models via Ollama, LM Studio, and Docker, so your code can stay on your machine. Under the hood it's a LangGraph deep agent, but you don't need to know that to use it.
Why it exists
Most coding agents optimize for autonomy — they act, and you find out afterward. CodeBuddy is built on the opposite conviction: an agent is only useful if you can trust it with your repository. So trust is wired into the architecture, not bolted on. Every file write goes through a review gate, and every shell command requires modal approval before it runs. You stay in the loop at exactly the moments that matter, and get out of the way everywhere else.
Purple steps are your checkpoints — nothing touches your files or shell until you say so.
How it's built
CodeBuddy is developed in the open by a single maintainer, with the full source, issue tracker, and roadmap public on GitHub. That means what you see is what runs — you can read the code that edits your code.
Who builds it
Have a question or want to get involved? Get in touch →