> Source: https://wexa.ai/docs/get-started/connect-an-ai-client

# Connect an AI client

Point Claude Desktop, Cursor or ChatGPT at your project's Wexa MCP server URL, so every Wexa tool appears in that client. You need a Wexa project and a role on it. [The Wexa MCP server](/docs/sdks-and-apis/mcp-clients) has the detail and the troubleshooting.

1. Find your project's connect URL, `https://fabric.wexa.ai/mcp/{projectId}`. Simple mode shows it in the **3 · Connect your agent** panel. Elsewhere, replace `{projectId}` with the `project_id` that `GET /v1/whoami` returns for your key.
2. Add the URL to your client. An interactive client runs OAuth in a browser.
   - **Claude Desktop:** Settings, **Connectors**, **Add**, **Add custom connector**, paste the URL, then sign in and approve.
   - **Cursor:** add the server below to `.cursor/mcp.json`, with an API key in `FABRIC_API_KEY` ([how to create one](/docs/get-started/authentication#create-an-api-key)).
   - **ChatGPT:** turn on developer mode, open **Settings → Connectors**, add a custom connector with the URL, then sign in and approve.

   ```json
   {
     "mcpServers": {
       "fabric": {
         "url": "https://fabric.wexa.ai/mcp/proj_abc123",
         "headers": { "Authorization": "Bearer ${env:FABRIC_API_KEY}" }
       }
     }
   }
   ```

3. Ask the client: "Using Wexa, tell me what my current scope and quota are." It should call the `docs` tool and name your identifiers. If it names the wrong project, the URL is wrong.

## Limits

| Limit | What happens |
| --- | --- |
| Scope | One connection reaches one project, never the platform at large. |
| Claude Desktop | Needs custom connectors on your plan; otherwise use [the local bridge](/docs/sdks-and-apis/mcp-clients#use-the-local-bridge), which works on any plan. |
| Cursor | Reads an API key from its environment; an administrator mints it. |
| ChatGPT | A developer-mode connector, not a deep-research source. |

**Add Wexa to Claude Desktop**

Open Settings, choose **Connectors**, click **Add**, then **Add custom connector**, and paste the URL. Sign in, confirm the organization, department and project, and approve.

Wexa's tools appear in the attachment menu at the bottom-left of the message box. If your plan has no custom connectors, [the local bridge](/docs/sdks-and-apis/mcp-clients#use-the-local-bridge) works on any plan.

**Add Wexa to Cursor**

Add a server to `.cursor/mcp.json` in your repository, or to `~/.cursor/mcp.json` for every project. This form reads an API key from Cursor's environment, so the file holds no secret:

```json
{
  "mcpServers": {
    "fabric": {
      "url": "https://fabric.wexa.ai/mcp/proj_abc123",
      "headers": {
        "Authorization": "Bearer ${env:FABRIC_API_KEY}"
      }
    }
  }
}
```

An administrator mints the key; see [Authentication](/docs/get-started/authentication#create-an-api-key). Enable Wexa under **Customize** in the sidebar. To use OAuth instead of a key, see [Cursor](/docs/sdks-and-apis/mcp-clients#cursor).

**Add Wexa to ChatGPT**

Turn on developer mode, open **Settings → Connectors**, add a custom connector, and give it a name and the URL. Sign in, confirm the project, and approve.

Wexa works as an ordinary developer-mode connector. It is not a deep-research source, because its search tools return code spans rather than documents. See [ChatGPT](/docs/sdks-and-apis/mcp-clients#chatgpt).

**More**

[Core ideas](/docs/get-started/core-ideas) explains why a connection is to one project. [How every tool call works](/docs/sdks-and-apis/how-every-tool-call-works) covers what each tool is called on your client's surface, what a call returns and how it is governed.