
Arbiter turns your existing AI into an autonomous developer interacting with your Github issues. It's an orchestration layer—handling Git, tests, and PRs while your AI writes the code. Works with Claude Code, Cursor, Ollama, and more. Use one, or mix and match for different tasks. 100% local-first CLI with a built-in web UI. Draft PRs by default so you review before merge. Runs your tests before committing. Comment on PRs to iterate automatically. Label issues, wake up to solved PRs.see more
Founder
Screenshots







About
Imagine having an tireless, expert developer on your team who never sleeps, understands your codebase intimately, and handles the tedious mechanics of software development so you can focus purely on the architecture and vision. That is exactly what Arbiter brings to your workflow. This isn't just another coding assistant; Arbiter is the sophisticated orchestration layer that empowers your favorite large language models—whether you prefer the reasoning power of Claude Code, the local flexibility of Ollama, or the specialized tooling within Cursor—to become truly autonomous agents within your GitHub ecosystem. Arbiter manages the entire development lifecycle initiated by an issue ticket. It handles the necessary Git operations, ensures code quality by running your existing tests before any changes are finalized, and automatically generates Pull Requests. This means you can wake up to fully drafted solutions, complete with necessary documentation and passing tests, ready for your final approval. It transforms the friction of context switching and manual repository management into a seamless, intelligent loop where your AI does the heavy lifting.
What truly sets Arbiter apart is its commitment to developer control and flexibility. While it automates the creation of code, it defaults to drafting Pull Requests, ensuring that human oversight remains the final checkpoint before merging into your main branch. Furthermore, its integration capabilities are vast, allowing you to mix and match different AI models for specialized tasks, leveraging the best tool for each specific challenge in your project. This tool understands the rhythm of development; it can automatically label issues, track progress, and even engage in iterative refinement by responding directly to comments you leave on the generated PRs. It’s designed to be a powerful, local-first Command Line Interface application, complete with a convenient built-in web UI, giving you transparency and access whether you are deep in the terminal or prefer a graphical overview.
Arbiter fundamentally changes the relationship between developer and AI. Instead of using an AI as a reactive tool that answers prompts, you are deploying it as a proactive, goal-oriented team member capable of taking an abstract requirement from a GitHub issue and delivering a tested, ready-to-review solution. This dramatically accelerates feature implementation and bug resolution cycles, freeing up your valuable cognitive load for higher-level problem solving. By automating the boilerplate of Git management, testing execution, and PR creation, Arbiter ensures that your development pipeline remains efficient, high quality, and always moving forward, making complex contributions feel almost instantaneous.