Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/neon-solutions/add-mcp/llms.txt

Use this file to discover all available pages before exploring further.

add-mcp is an open-source command-line tool that eliminates the friction of wiring MCP (Model Context Protocol) servers into AI coding agents. Instead of hunting down each agent’s config file, manually editing JSON or YAML, and restarting your editor, you run a single npx add-mcp command and the tool handles detection, configuration, and installation automatically across every agent you use.

Why add-mcp?

AI coding agents like Cursor, Claude Code, and VS Code all support MCP servers, but each one stores its configuration in a different file format and location. Adding the same server to three agents means editing three different files — and doing it again for every new project or team member. add-mcp treats all agents as a unified target, detects which ones are active in your current project or globally on your machine, and writes the correct config for each one.

Installation

No install step is required. Run add-mcp directly with npx:
npx add-mcp <url|package>
If you prefer to have the command available globally without npx, install it once:
npm install -g add-mcp

Supported Agents

add-mcp supports 15 AI coding agents. Pass the value from the --agent column to target a specific agent with the -a flag.
Agent--agent flag
Antigravityantigravity
Cline VSCode Extensioncline
Cline CLIcline-cli
Claude Codeclaude-code
Claude Desktopclaude-desktop
Codexcodex
Cursorcursor
Gemini CLIgemini-cli
GitHub Copilot CLIgithub-copilot-cli
Goosegoose
MCPortermcporter
OpenCodeopencode
VS Codevscode
Windsurfwindsurf
Zedzed
Several shorthand aliases are also accepted: codeium and cascade resolve to windsurf, cline-vscode resolves to cline, gemini resolves to gemini-cli, and github-copilot resolves to vscode.

Key Features

Smart agent detectionadd-mcp scans your current directory for agent config files (.cursor/mcp.json, .mcp.json, .vscode/mcp.json, and others) and pre-selects the agents it finds, so the default action is always the most sensible one for your project. Project and global scope — By default, servers are installed at the project level so the config can be committed alongside your code. Add the -g flag to install globally and make a server available across every project on your machine. Remote and local server support — Point add-mcp at a streamable HTTP or SSE URL to install a remote MCP server, or pass an npm package name to install a stdio-based local server via npx. Full shell commands and Node.js scripts are supported too. Registry search — The find command searches the add-mcp curated registry and the official Anthropic MCP registry so you can discover and install servers by keyword without knowing the exact package name or URL. Sync command — The sync command groups servers by URL or package name across all detected agents, normalizes their names, and propagates any missing entries — keeping every agent’s config consistent with a single command. Programmatic APIadd-mcp exports a TypeScript API from its main entry point, so you can integrate server installation into your own CLI tools or scripts without shelling out.

Quick Navigation

Quickstart

Install your first MCP server in under two minutes.

Add Command

Full reference for flags, transport types, and examples.

Agent Concepts

Understand project vs. global scope and how detection works.

Programmatic API

Use add-mcp as a library inside your own tools.

Build docs developers (and LLMs) love