Connect MCP clients

Your MCP server works with any client that supports the Model Context Protocol.

Audience	Developers and MCP client users configuring local AI tools
Prerequisites	A deployed MCP server and access to the target MCP client settings

Configuration

After deploying, you receive a JSON config snippet:

{
  "mcpServers": {
    "your-server": {
      "url": "https://appatools.com/mcp-studio/api/mcp/your-server"
    }
  }
}

Client-specific instructions

Cursor

Open Settings > MCP Servers, paste the JSON snippet, and restart Cursor.

Claude Desktop

Open Settings > Developer > Edit Config, add your server to mcpServers, and restart Claude Desktop.

Windsurf

Open the MCP configuration file, add the server entry, and restart Windsurf.

ChatGPT

ChatGPT on the web connects to MCP servers through developer mode, on Plus, Pro, Business, Enterprise, and Education plans. On a workspace plan, an admin enables developer mode for the workspace first.

1. Open Settings > Security and login and turn on Developer mode.
2. Go to Plugins and select the plus button. Enter a name and a short description.
3. Paste the server URL, choose No authentication, and create the app. ChatGPT lists the tools it found on the server.
4. In a new chat, open the plus menu, choose Developer mode, and select the app.

ChatGPT connects with OAuth or with no authentication and has no field for an access token, so connect a public server. Use a private server from Codex, Claude, or Cursor instead, which send the token with every request. After you change a server's tools, refresh the app from its details page in settings. ChatGPT asks before it runs a tool; you can remember your approval for the rest of a conversation.

ChatGPT on the web and Codex keep separate server lists, so an app created here does not appear in Codex. The full walkthrough is in How to connect an MCP server to ChatGPT, and OpenAI's reference is ChatGPT developer mode.

Codex

The ChatGPT desktop app, the Codex CLI, and the Codex IDE extension share one configuration, so add the server once.

• ChatGPT desktop app: open Settings > MCP servers > Add server, choose Streamable HTTP, paste the server URL, save, and select Restart.
• Codex CLI: run codex mcp add your-server --url https://appatools.com/mcp-studio/api/mcp/your-server.
• config.toml: add the server to ~/.codex/config.toml:

[mcp_servers.your-server]
url = "https://appatools.com/mcp-studio/api/mcp/your-server"
# Private servers only: the named variable holds your mcps_live_ token.
# bearer_token_env_var = "YOUR_SERVER_MCP_TOKEN"

For a private server, bearer_token_env_var takes the name of an environment variable, not the token. If you put the token there, Codex looks for a variable with that name, finds none, and skips the server. The ChatGPT desktop app may not see variables set in a terminal profile, so in the app you can send the header directly instead:

[mcp_servers.your-server]
url = "https://appatools.com/mcp-studio/api/mcp/your-server"
http_headers = { "Authorization" = "Bearer mcps_live_your_token" }

A server you added in ChatGPT's own settings (ChatGPT above) is configured separately. If it is missing from /mcp, add it to Codex as shown here.

Codex connects to servers when a session starts, so open a new session after adding one, then type /mcp to see the server and its tools. The full walkthrough is in How to connect an MCP server to Codex.

VS Code (with MCP extension)

Add the server URL to your MCP extension settings.

JetBrains (IntelliJ, WebStorm, PyCharm)

Configure through the MCP plugin settings.

Custom / SDK

Use the streamable HTTP endpoint URL (/api/mcp/{slug}) with any MCP-compatible SDK:

# Python
pip install mcp

# Node.js
npm install @modelcontextprotocol/sdk

Transports

Transport	Method	Endpoint
Streamable HTTP	POST	/api/mcp/{slug}

Every MCP client connects the same way.