How do I use Gemini CLI and Claude Code together?

Give both agents the same rules, the same tools and separate branches. Keep the rules in AGENTS.md: Claude Code reads it when the project has no CLAUDE.md (or through a CLAUDE.md with the line @AGENTS.md), and Gemini CLI reads it once context.fileName in .gemini/settings.json lists it. Register the same MCP servers in both. Give each agent its own git worktree, and pass work between them with a commit and a short handoff note. On a personal Google account, use Antigravity CLI instead of Gemini CLI: the same setup applies, and it reads AGENTS.md without a setting.

Can I still use Gemini CLI?#

It depends on how you sign in. On 18 June 2026, Gemini CLI stopped serving requests for Google AI Pro, Google AI Ultra and no-cost individual accounts, and Google moved those users to Antigravity CLI (Google's announcement). Gemini CLI still works with paid Gemini API keys and with Gemini Code Assist Standard or Enterprise licences.

Everything below works for either one. Where they differ, both are named.

Which rules file does each one read?#

For Gemini CLI, add this to .gemini/settings.json in the project:

{
  "context": {
    "fileName": ["AGENTS.md", "GEMINI.md"]
  }
}

Keep GEMINI.md only for notes that apply to Gemini alone, or drop it. One file of rules, committed and reviewed like code, means a convention you add for one agent reaches the other.

How do I give both the same tools?#

Both speak MCP, so a server you rely on (a database, an issue tracker, documentation search) can serve both. They keep separate lists:

claude mcp add --transport http --scope project docs https://docs.example.com/mcp
gemini mcp add --transport http --scope project docs https://docs.example.com/mcp

Claude Code writes project servers to .mcp.json; Gemini CLI writes them to .gemini/settings.json. Commit both so teammates get the same tools. Antigravity CLI keeps them in .agents/mcp_config.json. Keys belong in environment variables, never in these files.

How do I split the work?#

Run them side by side, not on top of each other:

1. One worktree each. git worktree add ../app-gemini -b gemini/search-index gives the second agent its own folder and branch, so neither edits a file the other has open.
2. One task each. Write down who owns what before either starts, in a task list both read.
3. Play to each one's strengths. A common split is one agent writing a feature while the other writes its tests, or one reviewing the other's pull request. A second model catches things the first one keeps missing.
4. Hand off on purpose. When work moves from one to the other, commit, then leave a note with the branch, the commit, what is done and the next step. The receiving agent reads the note, not the other agent's conversation.

Where does this break down?#

On one computer, a rules file, a worktree each and a task list are enough. It gets harder when:

• The agents belong to different people. Your Claude Code cannot see what a teammate's Gemini session decided an hour ago.
• The task list lives in a file. A claim in TASKS.md reaches the other agent only after a commit, a push and a pull.
• Questions pile up. Two agents in two terminals both stop for permission, and you find out when you look.

Where does Joxo fit?#

Joxo puts a team's agents in one project: Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Antigravity CLI and more, each on its own person's computer and plan. Only one agent can hold a task on the board, decisions reach every agent on its next turn, and a handoff carries the branch and the commit. Setup is the same line in every agent, pasted in the project folder:

Read https://joxo.ai/skill.md and follow it to set up Joxo in this project.

The agents page lists how well each agent has been checked with Joxo.

Frequently asked questions#

Can Claude Code call Gemini directly?#

Not out of the box. Claude Code can run any command, so it can call gemini -p "…" for a one-off answer, or reach Gemini through an MCP server. For real back-and-forth on a task, a shared task list and handoff notes work better than one agent calling the other.

Do I need both GEMINI.md and CLAUDE.md?#

No. One AGENTS.md can serve both, with Gemini CLI's context.fileName setting pointing at it. Keep the agent-specific files only for instructions that apply to one agent.

Does Antigravity CLI read my old Gemini CLI settings?#

Its rules still come from GEMINI.md and AGENTS.md, including ~/.gemini/GEMINI.md, but MCP servers, hooks and skills live in new places. Google's migration guide lists them.

Can a team mix Gemini, Claude Code and Codex?#

Yes. Each person keeps the agent and plan they already use. What they need in common is the rules file, the task list and a way to hand work over, which is what running several agents on one project covers.