MCP stack for Gemini CLI

Gemini CLI is the agent in your terminal. These three MCP servers work alongside it: one to run its work on machines you own, one to find more MCP servers, and one whiteboard you and Gemini look at together. Three gemini mcp add commands, all in user scope.

Add all three

Run these once. -s user writes them to ~/.gemini/settings.json so they work in every project; without it, gemini mcp add writes to the project's .gemini/settings.json (Gemini CLI MCP docs):

gemini mcp add -s user --transport http aicommander https://aicommander.dev/mcp
gemini mcp add -s user mcpfinder npx -y @mcpfinder/server
gemini mcp add -s user --transport http cnvs https://cnvs.app/mcp

Then run /mcp in Gemini CLI to see the servers and their tools. Gemini CLI does not connect MCP servers in a folder you have not trusted, so start it in a trusted project folder. Add only servers you trust: every MCP server you attach can act with the permissions you give it. If a server asks you to sign in, the first connect opens a browser sign-in.

ServerWhat it adds to Gemini CLITransportPrice
AI CommanderShell commands and detached jobs on your server, dev box, VPS or home PCRemote MCP (Streamable HTTP) + a host agent on each machineFree up to 10 machines; Pro $49/month
MCPfinderSearch the Official MCP Registry, Glama and Smithery; get install configLocal stdio via npx (Node.js 22.15+)Free, open source
cnvs.appA live whiteboard Gemini reads and draws on while you watchRemote MCPFree, no signup

AI Commander — Gemini CLI on your own machines

Install the host agent on each machine you want Gemini CLI to drive. It makes one outbound connection to aicommander.dev and accepts no inbound connections, so you open no ports. Gemini CLI can then run commands there and start detached jobs that keep running after the session ends, and check their status and logs later. There is no AI Commander app: Gemini CLI stays the UI. Prefer a local stdio bridge? gemini mcp add -s user -e AICOMMANDER_SERVER=https://aicommander.dev aicommander npx -y @aicommander/mcp. Full setup, including Antigravity CLI: AI Commander with Gemini.

Prompt — Gemini CLI
On my build-server, start a job in /srv/app named nightly-tests that runs npm test, then give me the jobId. I'll ask you for its status and logs later.

MCPfinder — let Gemini CLI find its own MCP servers

MCPfinder is an open-source (GitHub) MCP server that searches the Official MCP Registry, Glama and Smithery. Gemini CLI gets search_mcp_servers, get_server_details, get_install_config and browse_categories, so it can look up a server for the task and show you the config. Review what it suggests before you install it.

cnvs.app — one board for Gemini and you

cnvs.app is a free, no-signup, real-time whiteboard with an MCP endpoint at https://cnvs.app/mcp. Gemini CLI can read a board, add text, links and drawings, and manage kanban tasks on the same board you have open (cnvs.app MCP docs). The board ID is the key, so treat a board link like a password and keep secrets off shared boards.

The same stack for Claude Code, Codex, Grok, Cursor and Devin Desktop (Windsurf), plus PLAI.chat for second opinions from other models: Agent stack.

FAQ

Which MCP servers should I add to Gemini CLI first?
Add servers that give Gemini CLI something it cannot do alone. AI Commander gives it a shell and detached jobs on your other machines, MCPfinder lets it search the Official MCP Registry, Glama and Smithery for further servers and returns their install config, and cnvs.app gives it a whiteboard you and Gemini share. Add only servers you trust: every MCP server you attach can act with the permissions you give it.
Where does Gemini CLI keep its MCP servers?
In settings.json, under mcpServers: ~/.gemini/settings.json for user scope (gemini mcp add -s user), or .gemini/settings.json in a project, which is the default scope of gemini mcp add. A local stdio server uses command and args; a remote Streamable HTTP server uses httpUrl. Run /mcp inside Gemini CLI to see the servers and their tools.
Is AI Commander an alternative to Gemini CLI?
No. AI Commander is a remote harness that works alongside Gemini CLI, Claude Code, Codex, Cursor and Grok. It has no chat or app of its own: Gemini CLI stays the UI, and AI Commander lets it run commands and detached jobs on machines you own.
Can Gemini CLI keep working on my server after I close the session?
Yes, with a detached job. Ask Gemini CLI to start an AI Commander job on the machine; the job keeps running after the session ends and you read its status and logs later. The host agent dials out, so the server needs no inbound ports.
What does this cost?
AI Commander is free for up to 10 machines; Pro is $49 per month and adds file transfer and more machines. MCPfinder and cnvs.app are free. These are separate products from the Coder AI team with separate pricing; there is no bundle.

Start with your own machine

Install the host agent on a machine you own, run gemini mcp add -s user --transport http aicommander https://aicommander.dev/mcp, and ask Gemini CLI for a real job on that box.

Install the host agent Gemini setup AI Commander pricing

Disclosure: AI Commander, MCPfinder and cnvs.app are made by the Coder AI team. Prices as published on 2026-10-20. Gemini is a trademark of Google LLC; AI Commander is not affiliated with, endorsed by, or sponsored by Google.