add-mcp CLI
add-mcp writes an MCP server into the configs of 25 agents with one command: Claude Code, Codex, Cursor, VS Code, Zed and more
- 321 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Edits agent config files, including global ones in the home directory
- Values from --env and --header are written into the agent config, and a --gitignore flag exists for shared repositories
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/add-mcpDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Edits agent config files, including global ones in the home directory
- Values from --env and --header are written into the agent config, and a --gitignore flag exists for shared repositories
This is third-party code. Review the repository files before installing.
What it does
add-mcp is a command-line utility that writes an MCP server into the config files of several agents in one call. The input is a remote server URL, an npm package name or a launch command, for example npx add-mcp https://mcp.context7.com/mcp. The tool detects agents from project files (.cursor/, .vscode/, .mcp.json), offers a selection and writes an entry in each agent's format. By default the install goes into the project, and with -g into the home directory. The flags --env, --header and --transport set environment variables, headers and the connection type (http or sse). find searches servers in the add-mcp registry, list shows installed ones, remove deletes, and sync aligns names across agents. --auto-approve turns on tool auto-approval in Codex and Claude Code. A programmatic upsertServer function exists for embedding in your own tools.
Who it is for. Developers who connect the same MCP servers to several agents and do not want to edit JSON and TOML by hand.
Good fit when
- A server has to go into Claude Code, Cursor and Codex at once with a single command
- The repository holds a shared .mcp.json and the team wants servers installed the same way
- You search for a server by keyword: npx add-mcp find vercel
- Names of the same servers drifted apart across agents and sync brings them back in line
Not a fit when
- The agent is not among the 25 supported ones and does not read MCP configs
- The server is already wired in through an agent plugin and a second config entry is pointless
- You need a server catalog with ratings and security vetting: add-mcp writes whatever you point it at
Example request
Add the Context7 MCP server to Claude Code and Cursor for this project with add-mcpLimitations
- Not every agent understands every field: --timeout, --scopes and --bearer-token-env apply selectively, and for other agents the tool prints a warning and drops the field.
- Pi needs the pi-mcp-adapter, and Kimi Code reads a project-level .kimi-code/mcp.json only after you trust the folder.
- A running session needs a config reload: /reload-mcp in Hermes Agent, /mcp reload in fx, /reload in Pi.
- fx ignores project files: it reads servers only from ~/.fx/mcp.json.
How to disable. Remove a server with npx add-mcp remove and the server name, or by hand from the agents' config files. The registry saved on the first search lives in ~/.config/add-mcp/config.json.
FAQ
How do I add an MCP server to Claude Code and Cursor with one command?
Run npx add-mcp server-url -a claude-code -a cursor. Without -a the tool detects agents from project files and shows a selection.
How does add-mcp differ from mcpm?
add-mcp only writes a server into agent configs, while mcpm keeps a registry and updates servers. The first suits a one-off install, the second maintaining a set.
How do I see what is already installed?
npx add-mcp list shows servers across all detected agents. list-agents prints the supported agents with their scopes, project and global.
How do I install an MCP server globally?
Add the -g flag: npx add-mcp https://mcp.example.com/mcp -g -a claude-code -y. Without it the entry lands in the current project's files.
How do I pass an API key to a server?
For a local server use --env "API_KEY=value", for a remote one --header "Authorization: Bearer ...". Values land in the agents' config files, so add --gitignore for a shared repository.
Discussion
No questions or reviews yet.
Related
CLI tools1MCP Agent
1mcp-agent
One MCP server instead of many: the agent merges connected servers and gives them to a client through a single address. A beta version
An MCP server that lets an agent read files, search code and run commands in a workspace
The openmcp CLI and a VS Code extension for debugging MCP servers: tools, prompts and resources, chat with a model, test cases
A plugin with skills for building TypeScript MCP servers and MCP apps: the agent scaffolds, edits, debugs and verifies them
A command-line debugger for MCP servers from Giant Swarm: a REPL for tools, OAuth 2.1 and an MCP server mode for Cursor
A lightweight CLI to discover, inspect and call MCP server tools straight from the shell
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| add-mcp (this card) | no | 321 | ||
| MCPM | no | 1,021 | an MCP package manager and registry that also updates servers | |
| mcp-installer | no | not required | 1,535 | an MCP server that installs other servers from npm and PyPI on the agent's request |
README in short
The README opens with an install command and a Context7 example, then covers registry search, a table of 25 agents with project and global file paths, and quirks of individual clients. Next come install scopes, agent auto-detection, the find, list, remove and sync commands, flag examples and a table of fields that not every agent supports.