WhatsApp MCP server for Cursor, VS Code, Gemini CLI and Codex

If you build with an AI coding assistant, you can give it WhatsApp: send a test notification while you develop, submit the template your app needs, check delivery status — or have it wire WhatsApp into your app through the REST API, without leaving the editor. SendFromChat is a hosted MCP server for Meta's official WhatsApp Cloud API, so there's nothing to run locally.

Why a hosted MCP server

You could call Meta's Graph API directly, but you'd be rebuilding the parts that take time: token storage, template validation, the 24-hour window rules, webhook handling, pacing under messaging limits, opt-out tracking. A hosted server gives your assistant all of that as tools, over streamable HTTP — the transport every current MCP client supports.

Two ways to connect:

• Sign in (recommended): use https://sendfromchat.com/api/mcp. Clients that support MCP sign-in open a browser where you sign in to SendFromChat and click Allow — no key in any file.
• Private connector URL: for clients without sign-in, use https://sendfromchat.com/api/mcp/YOUR_KEY, generated on the Connect your AI assistant card in the dashboard. Or use the plain URL with an Authorization: Bearer YOUR_KEY header, if your client supports headers.

The examples below use the plain URL. If your client doesn't offer to sign in, swap in your private URL.

Cursor

Settings → MCP → Add new MCP server, or edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "sendfromchat": { "url": "https://sendfromchat.com/api/mcp" }
  }
}

VS Code (GitHub Copilot, Agent mode)

Command Palette → MCP: Add Server → HTTP, paste the URL, name it sendfromchat. Or in .vscode/mcp.json:

{
  "servers": {
    "sendfromchat": { "type": "http", "url": "https://sendfromchat.com/api/mcp" }
  }
}

Windsurf

In ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "sendfromchat": { "serverUrl": "https://sendfromchat.com/api/mcp" }
  }
}

Gemini CLI

In ~/.gemini/settings.json:

{
  "mcpServers": {
    "sendfromchat": { "httpUrl": "https://sendfromchat.com/api/mcp" }
  }
}

Start gemini and run /mcp to confirm it's connected.

Codex

In ~/.codex/config.toml:

[mcp_servers.sendfromchat]
url = "https://sendfromchat.com/api/mcp"

Claude Code

claude mcp add --transport http sendfromchat https://sendfromchat.com/api/mcp

Then run /mcp, choose sendfromchat → Authenticate, and allow it in the browser. Ask it to "add WhatsApp order confirmations to this app" and it writes the integration with your real templates, saving the API key straight into .env.local — the key never appears in the chat.

Anything else

Any MCP client that supports remote servers over streamable HTTP works: add a server, choose HTTP, paste the URL. If your client only supports local (stdio) servers, use a small bridge such as mcp-remote.

Useful developer prompts

• "Send the hello_world template to my number to check the integration."
• "Create a UTILITY template for password-reset codes in English and Spanish and tell me when they're approved."
• "Is my webhook receiving delivery receipts? Show the last five messages and their status."
• "Upload test-users.csv as a list called staging_testers."
• "Add WhatsApp order confirmations to this app, with a webhook for replies."

The full list of tools is in the MCP tools reference.

Security notes

• Prefer signing in: nothing secret ends up in a config file, and you can disconnect any client under Settings → Connected apps.
• A private connector URL grants full access to your workspace. Keep configs containing it out of version control (use a user-level config, not a repo file), or use the header form with an environment variable where your client supports it. Regenerating it in the dashboard immediately disables the old one.
• API keys your assistant sets up for an app are saved straight into .env.local or your host, never shown in the conversation.