> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ringee.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect a client

> Get your Ringee MCP URL and wire it into Claude, ChatGPT or Cursor

## Get your MCP URL

While signed in to Ringee, call:

```http theme={null}
GET /api/mcp/connection-info
```

```json theme={null}
{
  "mode": "organization",
  "userId": "a91c…",
  "organizationId": "org_2k7yX…",
  "url": "https://api.ringee.io/api/mcp/a91c…/org_2k7yX…/sse"
}
```

The URL matches your currently active workspace:

| Mode         | URL shape                                                     |
| ------------ | ------------------------------------------------------------- |
| Personal     | `https://api.ringee.io/api/mcp/<userId>/sse`                  |
| Organization | `https://api.ringee.io/api/mcp/<userId>/<organizationId>/sse` |

<Note>
  Either URL works from any client — the assistant can move between workspaces at runtime with `switch_workspace`. See [Workspaces](/mcp/workspaces).
</Note>

<Warning>
  This URL is a credential. Anyone holding it can read and modify that workspace's contacts, calls and sessions. Do not share it, and generate connection info again if you believe it leaked.
</Warning>

## Transport

The endpoint speaks MCP over **SSE**:

| Method | Path                                    | Purpose                |
| ------ | --------------------------------------- | ---------------------- |
| `GET`  | `/api/mcp/:userId/sse`                  | Open the event stream  |
| `POST` | `/api/mcp/:userId/messages?sessionId=…` | Send JSON-RPC messages |

The org-scoped form adds `/:organizationId` before `/sse` and `/messages`. The `messages` endpoint requires the `sessionId` the stream handed you.

## Claude Desktop and claude.ai

<Tabs>
  <Tab title="claude.ai">
    1. Open **Settings → Connectors → Add custom connector**.
    2. Paste your MCP URL.
    3. Save, then start a chat and ask Claude to list your Ringee contacts.

    The server must be reachable from the public internet — `localhost` will not work.
  </Tab>

  <Tab title="Claude Desktop">
    Add the server to your MCP configuration:

    ```json theme={null}
    {
      "mcpServers": {
        "ringee": {
          "url": "https://api.ringee.io/api/mcp/<userId>/sse"
        }
      }
    }
    ```

    Restart Claude Desktop and confirm the Ringee tools appear in the tools list.
  </Tab>

  <Tab title="Claude Code">
    ```bash theme={null}
    claude mcp add --transport sse ringee https://api.ringee.io/api/mcp/<userId>/sse
    ```

    Or install the full Ringee plugin, which bundles the MCP server and the `/ringee` skill commands — see [Apps and skills](/mcp/apps).
  </Tab>

  <Tab title="Cursor">
    Add the server under **Settings → MCP**, using the same SSE URL. Cursor discovers the tool list automatically once the stream connects.
  </Tab>
</Tabs>

## ChatGPT

ChatGPT connectors require OAuth, so they use a different, id-free endpoint:

```
https://api.ringee.io/api/mcp/chatgpt/sse
```

This endpoint resolves the caller from a verified bearer token instead of trusting ids in the URL, so every end user reaches only their own data. It publishes RFC 9728 protected-resource metadata for discovery:

```http theme={null}
GET /api/mcp/chatgpt/.well-known/oauth-protected-resource
```

Self-hosted deployments configure the authorization server with:

| Variable                          | Purpose                                                                  |
| --------------------------------- | ------------------------------------------------------------------------ |
| `MCP_OAUTH_AUTHORIZATION_SERVERS` | Your Clerk issuer (comma-separated if more than one)                     |
| `MCP_OAUTH_SCOPES`                | Scopes advertised — defaults to `ringee:use`                             |
| `OPENAI_APPS_CHALLENGE_TOKEN`     | Domain-verification token served at `/.well-known/openai-apps-challenge` |

<Info>
  OpenAI issues a **distinct** verification token per origin. The backend (`api.ringee.io`) and a separately deployed ChatGPT MCP server each need their own.
</Info>

## Verify the connection

The fastest check is the [`ringee` CLI](/cli/overview), which connects to the same server:

```bash theme={null}
export RINGEE_MCP_URL="https://api.ringee.io/api/mcp/<userId>/sse"
ringee config check
```

It connects and prints every tool the server exposes. If that works, any MCP client will too.

You can also use the MCP Inspector:

```bash theme={null}
npx @modelcontextprotocol/inspector
```

## Self-hosting

Point clients at your own backend instead:

```
https://ringee-api.example.com/api/mcp/<userId>/sse
```

The URL is built from `BACKEND_URL`, so make sure that variable is the **public** URL of your API — see the [Configuration reference](/configuration/reference).

## Next steps

<CardGroup cols={2}>
  <Card title="Tool reference" icon="list" href="/mcp/tools">
    Everything the assistant can call
  </Card>

  <Card title="Safety" icon="shield-check" href="/mcp/safety">
    How destructive and credit-spending actions are guarded
  </Card>
</CardGroup>