# How to use Claude Code, Codex and Gemini CLI together

Using several coding agents together means giving each the tasks it suits, on the same projects,
and comparing what they do. Each vendor's agent is its own command-line tool with its own sign-in,
settings and commands. Moontors, an agentic development environment (ADE), runs all three, Claude
Code, Codex and Gemini CLI, side by side on one board, each as its vendor ships it.

## Sign in to each

Each agent signs in with its own vendor:

- Claude Code: a Claude subscription in the browser, or an Anthropic API key;
- Codex: a ChatGPT sign-in in the browser or with a device code, or an OpenAI API key;
- Gemini CLI: a Gemini API key. Google no longer lets personal accounts sign in to it through the
  browser; an organization account may still.

Moontors installs each agent the first time you use it, and keeps its sign-in apart from the agent
you use outside Moontors. See [Signing in](/docs/sign-in/).

## Give each agent its tasks

Choose **Create session** and pick the agent as the provider, with its model. Sessions of different
agents can work on the same project at once:

- in the same folder, for tasks that touch different files;
- each in its own worktree, for tasks that might touch the same files. See
  [Worktrees](/docs/worktrees/).

Every card names its agent and model, and you can filter the board by project to see every
agent's work on one repository.

## Compare their work

To try the same task with two agents, create a worktree session for each, with the same prompt.
Each works on its own branch, and each session's Git view shows what it changed. See
[Git and Files](/docs/diff-and-files/#git).

## What stays each agent's own

- Its settings, MCP servers, skills, plugins, hooks and instruction files (`CLAUDE.md`, `AGENTS.md`
  or `GEMINI.md`) are used as they are, a plugin with all of its parts. See
  [Configuration](/docs/configuration/#plugins) for how each agent takes them, and
  [Known limits](/docs/known-limits/#configuration) for what Moontors can't switch or hasn't tested.
- Its own slash commands work in its sessions, beside Moontors' own. Each agent's commands
  reference says which do what:
  [Claude Code](/docs/commands/claude/), [Codex](/docs/commands/codex/) and
  [Gemini CLI](/docs/commands/gemini/).
- Its account's usage and limits are its own.

## Questions

### Can the three agents work on the same repository at once?

Yes. Give each session its own worktree when their tasks might touch the same files, and each
agent works on a branch of its own.

### Do I need an account with each vendor?

Yes. Each agent signs in with its own vendor: a Claude plan or an Anthropic API key, a ChatGPT
plan or an OpenAI API key, and a Gemini API key or, where Google still allows it, an organization
account's browser sign-in. Moontors itself needs no account. See [Signing in](/docs/sign-in/).

### Can every agent do the same things in Moontors?

Not quite. Some commands work with some agents only, and Gemini CLI does less than the others in
Moontors; [Known limits](/docs/known-limits/#sessions) lists each.

## Related

- [What is an agentic development environment (ADE)?](/docs/what-is-an-ade/)
- [How to run several Claude Code sessions in parallel](/docs/how-to/run-claude-code-sessions-in-parallel/)
- [A Codex app alternative with Claude Code and Gemini CLI](/docs/compare/chatgpt-codex/)
- [An Orca alternative with Gemini CLI in structured views](/docs/compare/orca/)
- [An Emdash alternative with Gemini CLI and a background service](/docs/compare/emdash/)