For AI agents: the index of this site is https://dejo.run/llms.txt, and every page is also served as markdown (add .md to its address, or send Accept: text/markdown).

dejo

MCP config generator for Claude Code, Codex, Cursor, VS Code and Claude Desktop

The short answer: describe your MCP server once (its name, the command that starts it, its arguments and any environment variables) and copy the exact config for each AI client. Every format below was checked against the client's own documentation on 10 November 2026.

Start from
Environment variables
Run in a terminalany folder
claude mcp add --scope user --transport stdio dejo -- dejo-mcp

--scope user: every project; project: writes .mcp.json for the team; local: only you, only this project.

Or add to .mcp.json.mcp.json in the project root
{
  "mcpServers": {
    "dejo": {
      "command": "dejo-mcp"
    }
  }
}

It runs in your browser; nothing you type is sent anywhere. It covers local servers, the kind started with a command (stdio). It starts with Dejo, a browser MCP that remembers websites, filled in; press "Any server" to start from scratch.

Where each client keeps its MCP config

ClientHow to add a serverTop-level key
Claude Codeclaude mcp add in a terminal, or .mcp.json in the projectmcpServers
Codexcodex mcp add, or ~/.codex/config.toml[mcp_servers.NAME] table
Cursor~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project)mcpServers
VS Code.vscode/mcp.json in the workspaceservers
Claude Desktopclaude_desktop_config.jsonmcpServers

Claude Code

The command is the simplest way. Everything after -- is the command that starts the server, passed through untouched. Environment variables go before the server's name with --env KEY=value, and another option must sit between --env and the name, which is why the generator puts --transport stdio there.

--scope decides where the server is saved: user for every project, project for a .mcp.json file the whole team shares, and local, the default, for you in this project only.

Codex

Codex reads ~/.codex/config.toml, or a .codex/config.toml inside a trusted project. Each server is its own table, [mcp_servers.NAME], with command and an optional args list. Environment variables go in a separate [mcp_servers.NAME.env] table. Codex waits 10 seconds for a server to start unless you set startup_timeout_sec.

Cursor

Cursor reads ~/.cursor/mcp.json for servers you want everywhere and .cursor/mcp.json for one project. Its documentation lists type as required, so the generator includes "type": "stdio".

VS Code

VS Code reads .vscode/mcp.json in the workspace, or a mcp.json in your user profile. It is the one client here whose top-level key is servers, not mcpServers. A config copied from another client fails until that key is renamed.

Claude Desktop

Claude Desktop reads claude_desktop_config.json:

Quit Claude Desktop completely and open it again after you save, so it loads the new server.

Sources

FAQ

Where is the Claude Code MCP config file?

Claude Code stores servers you add with `claude mcp add` for you. A server added with `--scope project` goes into `.mcp.json` in the project root, under a top-level `mcpServers` key, so the whole team gets it. `--scope user` makes it available in every project, and `local` (the default) keeps it to you in this project.

How do I add an MCP server to Codex?

Either run `codex mcp add NAME -- COMMAND ARGS`, or add a `[mcp_servers.NAME]` table to `~/.codex/config.toml` with `command` and `args`, and environment variables in a `[mcp_servers.NAME.env]` table.

Why does my VS Code MCP config not work when I copy it from Cursor?

VS Code uses a top-level `servers` key, while Cursor, Claude Code and Claude Desktop use `mcpServers`. Each VS Code entry also says `"type": "stdio"` for a local server.

Is anything I type sent anywhere?

No. The generator runs entirely in your browser. Nothing you type, including API keys in environment variables, leaves the page.