I built a terminal that puts Claude Code, Codex and Gemini in the same conversation. Here's what I learned.
I kept switching between terminal windows — Claude Code in one, Codex in another. Copy-pasting context between them manually. I was the middleman.
So I built Fixy Code to remove myself from that loop.
What it is
Fixy Code is an open source terminal tool that lets Claude Code, Codex and Gemini share one conversation thread. You address each agent directly with @mentions. They see each other's output and can review each other's work.
@claude review this function
@codex do you agree?
@all build the auth middleware
The @all command triggers a collaboration loop — agents discuss, plan, execute in batches, and review each other's output before anything gets committed.
Current status
Launched 3 days ago. Published on npm. Open source, MIT licensed. Free tier available. Zero paying users so far — being honest.
What I'd do differently
I built too many features before showing anyone. If I started over I'd ship @claude and @codex in one thread and nothing else. Validate the core before building everything else.
The honest risk
Claude Managed Agents launched last week. OpenAI's Codex plugin for Claude Code launched two weeks ago. The space is moving fast. The window for a neutral cross-company layer may be shorter than I'd like.
Try it
npm install -g @fixy/code
Run fixy from inside any git repo.
https://github.com/fixy-ai/fixy-code
Curious if anyone else has tried running multiple AI coding tools on the same codebase. What did you find?
