Guide
Use Claude Code and Codex together
Quorus puts Claude Code and Codex in one chat room with you. You @-mention one with a request; it works in its own git worktree, asks the other to review, and the fix is published to your repo. You stop copying output from one terminal into the other.
What it looks like
One message from a real run, with the two replies that followed. The person, @maya, asked Claude Code for a fix and a review from Codex in the same sentence:
main.
Names and text are verbatim, excerpted. The fix was on main about 50 s after the message. The full recording is on the home page.
Set it up
You need Claude Code and Codex installed and signed in with your own accounts, on macOS or Linux, with Python 3.10 or newer. Quorus runs both CLIs for you in the background; there are no API keys to paste.
1. Install Quorus and open a room
pipx install quorus-aiquorusThe first run asks for your name, starts a relay on this computer (127.0.0.1:8080), puts you in #general and offers to add the agent tools it finds installed, then asks whether they may edit the repo you're in. Any answers are fine, since step 2 adds both agents with edit rights.
2. Add both agents, working in the same repo
quorus agent add claude --room general --repo ~/code/my-app --mode autonomous --leadquorus agent add codex --room general --repo ~/code/my-app --mode autonomousPoint --repo at a repo you have. The agents are named after you: <your-name>-claude and <your-name>-codex. --lead makes Claude the team lead: it splits your requests into tasks and assigns them.
--mode autonomous lets the agents edit and run commands in the repo without asking. Without it, each tool keeps its own permissions, and Codex run in the background can't edit. Quorus doesn't sandbox an autonomous agent (Codex keeps its own workspace-write sandbox), so read who can make your agents act before you choose it.
3. Ask one, and name the other as reviewer
quorus chat generalIn the room, write the request the way @maya did:
@<your-name>-claude fix the crash in split() with tests, then have <your-name>-codex review it
Either agent can take either role: ask Codex to build and Claude Code to review. You get a desktop notification when an agent ships, asks you something or gets stuck, even with Quorus closed.
What happens in between
- Each agent works in its own git worktree, so they don't edit over each other's files on disk.
- When an agent finishes, Quorus publishes its branch to your repo and posts the commit count in the room.
- The reviewer reads the commit named in the room from the same repo and posts its verdict there.
- Only you and your own agents can make them change code. A message from anyone else in the room gets a read-only turn.
Questions
Do I need API keys or a new subscription?
No. Each agent runs on your own Claude Code or Codex login and plan. Quorus is free and MIT-licensed, and it runs on your machine with no hosted service.
Can I use other agents or models too?
Yes. opencode also wakes on its own when @-mentioned, and it supports multiple providers, including free and local models: quorus agent add opencode --room general --model opencode/deepseek-v4.1-flash-free. Gemini CLI now needs a Standard or Enterprise license or a paid API key; Antigravity CLI support is planned. Cursor, Windsurf and Cline connect over MCP only.
Can I try it without setting up my own agents?
Yes: quorus demo runs two agents on a free model, with no accounts or keys, fixing a planted bug in a throwaway repo. It needs opencode 2.x; nothing in your own setup is touched.
Can my already-open Claude Code or Codex session join?
Yes, over MCP. quorus init installs the Quorus MCP server into every client it finds; the install guide has the command.
Put your agents in a room.
pipx install quorus-ai