Bridge MCP — Claude Code MCP server for Windows control and setup
Bridge MCP connects an AI agent to a Windows desktop so it can operate apps, interact with the screen and automate Chrome. Use it to hand desktop tasks to an agent.
claude mcp add bridge-mcp -- uvx bridge-mcpWhat Bridge MCP does
The MCP server relays requests to a local Windows agent. An AI client can launch or switch apps, use the mouse and keyboard, inspect interface elements, capture the screen and control Chrome. Browser workflows use Playwright, whose browsers are installed on first use.
The local agent needs Python 3.10+ and the packages listed in requirements-local.txt. Start it with python local_agent.py and keep the terminal open; it listens on port 8006. Risky actions, including file writes and shell-script execution, are blocked unless approved. Choosing to approve every future command turns Safe Mode off, so the agent then has fewer approval checks.
What it helps with
- control application windows through an AI client
- click interface elements and enter text on Windows
- capture the desktop and inspect UI elements
- automate Chrome browsing tasks
- run PowerShell commands and handle files
Example prompts for the agent
- Open the settings window in the requested app.
- Find the submit button on this page and click it.
- List the open windows and capture a screenshot.
How to install Bridge MCP
Platform CLI
claude mcp add bridge-mcp -- uvx bridge-mcpuvx runs the command named after the package; if the README names another one, use uvx --from bridge-mcp <command>
Platform CLI
codex mcp add bridge-mcp -- uvx bridge-mcpuvx runs the command named after the package; if the README names another one, use uvx --from bridge-mcp <command>
Config file
{
"mcpServers": {
"bridge-mcp": {
"command": "uvx",
"args": [
"bridge-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from bridge-mcp <command>
Config file
{
"servers": {
"bridge-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"bridge-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from bridge-mcp <command>
Platform CLI
gemini mcp add bridge-mcp uvx bridge-mcpuvx runs the command named after the package; if the README names another one, use uvx --from bridge-mcp <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 Bridge MCP
How do I install Bridge MCP in Claude Code?
Run claude mcp add bridge-mcp -- uvx bridge-mcp. To run the local agent, install packages with pip install -r requirements-local.txt, then start it using python local_agent.py.
What does Bridge MCP do?
It lets an AI client send desktop commands to a Windows agent, including app control, keyboard and mouse input, screen inspection and Chrome automation.
What is needed to run the Bridge MCP local agent?
Use Windows with Python 3.10+. Install the packages from requirements-local.txt, then run python local_agent.py and keep it running in a terminal.
