Open-source tool 'Errorta' lets you run multiple AI coding CLIs as one team
A developer built an open-source CLI (with a UI too) called Errorta to solve the hassle of juggling different — like Claude Code, Codex CLI, and Cursor CLI — on the same project. Inside Errorta, you set up a 'team' of models and assign each one a role: for example, a Project Manager (PM) role, developer roles, a tester role, and a reviewer role. The PM-role model looks at the overall goal (the 'north star') and breaks it into tasks for the developer-role models to implement.
Testers and reviewers then check the finished work, and if something's wrong, they can reject it and send it back to the PM to redistribute. The is available via Homebrew, and it can connect to Claude CLI, Codex CLI, Cursor CLI, various AI APIs, and local LLMs. The creator's favorite setup so far uses Opus as PM, as developer, and Sonnet for both testing and reviewing.
The project's GitHub repo is named errorta_app.
Key points
- Errorta is an open-source CLI/UI that lets you combine Claude CLI, Codex CLI, Cursor CLI, AI APIs, and local LLMs into one team on a single project
- You can assign roles like Project Manager (PM), developer, tester, and reviewer to different models
- The PM model breaks the overall goal into tasks for developers; testers and reviewers check the work and can reject it back to the PM
- The creator's favorite combo so far: PM=Opus, developer=, tester/reviewer=Sonnet
- Currently an , installable via Homebrew; GitHub repo is errorta_app