Agent Intern — Claude Code MCP server: what it does, how to install
Connects external CLI tools like Gemini, Codex, and Copilot as sub-agents inside Claude Code. Offload routine edits and preserve your primary token limits.
claude mcp add agent-intern -- uvx agent-internWhat Agent Intern does
The server exposes existing command-line AI tools as callable functions inside Claude Code. It routes tasks to Gemini via agy, ChatGPT via codex, GitHub Copilot, or Cursor. You do not need dedicated server API keys because the bridge relies on the active logins from your installed CLI sessions. If you lack paid subscriptions, running the opencode-ai utility provides free models without credentials.
Parallel execution handles multiple errands at once. Tooling like agent_swarm runs concurrent tasks across different engines, while built-in routines request second opinions from competing model families to spot blind spots in diffs. Gemini handles graphic creation, saving files straight into the workspace.
The server will not launch via uvx unless you have uv installed locally. You also need at least one supported CLI pre-installed and authenticated on your machine. Passing watch=true pops open a live step-by-step progress monitor.
What it helps with
- hand off diff reviews to competing model families
- generate graphics via Gemini directly into project folders
- process batches of source files concurrently using swarms
- cut Claude token burn by running chores through secondary CLIs
Example prompts for the agent
- Ask Gemini to draw a logo for the docs and place it into
assets/. - Inspect the latest git diff using Copilot and highlight disagreements.
- Run a parallel swarm to summarize every file inside
src/handlers.
How to install Agent Intern
Platform CLI
claude mcp add agent-intern -- uvx agent-internuvx runs the command named after the package; if the README names another one, use uvx --from agent-intern <command>
Platform CLI
codex mcp add agent-intern -- uvx agent-internuvx runs the command named after the package; if the README names another one, use uvx --from agent-intern <command>
Config file
{
"mcpServers": {
"agent-intern": {
"command": "uvx",
"args": [
"agent-intern"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from agent-intern <command>
Config file
{
"servers": {
"agent-intern": {
"type": "stdio",
"command": "uvx",
"args": [
"agent-intern"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from agent-intern <command>
Platform CLI
gemini mcp add agent-intern uvx agent-internuvx runs the command named after the package; if the README names another one, use uvx --from agent-intern <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Agent Intern
How to install Agent Intern in Claude Code?
Run claude mcp add agent-intern -- uvx agent-intern. Ensure uv is installed on your system before running it.
What does Agent Intern do?
It bridges Claude Code to external coding CLIs, turning Gemini, Codex, Copilot, or Cursor into delegatable workers for code review, image generation, and batch jobs.
Does Agent Intern require its own API keys?
No. It uses the existing sessions of your authenticated CLI tools, or runs opencode-ai for zero-credential tasks.
