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

# Client Setup

> Connect Claude, Claude Code, Claude Desktop, Cursor, ChatGPT, or any other MCP client to the Blockworks MCP server.

Every client connects to the same hosted endpoint and authenticates through the same browser OAuth flow. Only the
place you paste the URL changes.

<Card title="MCP server URL" icon="link">
  ```
  https://mcp.blockworks.com/mcp
  ```
</Card>

<Tabs>
  <Tab title="Claude">
    Add Blockworks as a custom connector in Claude on the web or in the desktop app.

    <Steps>
      <Step title="Open connector settings">
        Go to [claude.ai/settings/connectors](https://claude.ai/settings/connectors).
      </Step>

      <Step title="Add a custom connector">
        Click **Add custom connector**.
      </Step>

      <Step title="Enter the server URL">
        Use `https://mcp.blockworks.com/mcp`.
      </Step>

      <Step title="Connect and authenticate">
        Click **Connect**, then enter your Blockworks API key when the OAuth window prompts you.
      </Step>
    </Steps>

    Blockworks then appears in the connectors list for your chats. Enable it in a conversation to let Claude call
    its tools.
  </Tab>

  <Tab title="Claude Code">
    Add the server with one command:

    ```bash theme={null}
    claude mcp add --transport http blockworks https://mcp.blockworks.com/mcp
    ```

    By default this registers the server for your user across projects. Add `--scope project` to write it into the
    project's `.mcp.json` so your teammates pick it up too.

    Equivalent JSON, if you prefer to edit the config directly:

    ```json .mcp.json theme={null}
    {
      "mcpServers": {
        "blockworks": {
          "type": "http",
          "url": "https://mcp.blockworks.com/mcp"
        }
      }
    }
    ```

    Then, inside Claude Code:

    * Run `/mcp` to see the server and start the browser OAuth flow if it has not run yet.
    * Confirm the server shows as connected and its tools are listed.

    <Tip>
      `claude mcp list` shows every configured server and its connection state, which is useful for confirming
      the entry landed in the scope you expected.
    </Tip>
  </Tab>

  <Tab title="Claude Desktop">
    Claude Desktop connects to hosted servers through the same custom connector mechanism as claude.ai.

    <Steps>
      <Step title="Open settings">
        In Claude Desktop, open **Settings** and go to **Connectors**.
      </Step>

      <Step title="Add a custom connector">
        Click **Add custom connector** and enter `https://mcp.blockworks.com/mcp`.
      </Step>

      <Step title="Authenticate">
        Click **Connect** and enter your Blockworks API key in the browser window that opens.
      </Step>
    </Steps>

    <Note>
      Custom connectors are the supported path for remote servers. The `claude_desktop_config.json` file is for
      local `stdio` servers and is not the right place for a hosted HTTP endpoint.
    </Note>
  </Tab>

  <Tab title="Cursor">
    Add Blockworks to your Cursor MCP configuration: `~/.cursor/mcp.json` for every project, or `.cursor/mcp.json`
    inside a single project:

    ```json .cursor/mcp.json theme={null}
    {
      "mcpServers": {
        "blockworks": {
          "url": "https://mcp.blockworks.com/mcp"
        }
      }
    }
    ```

    Reload Cursor, open **Settings** and then **MCP**, and confirm the server is listed. Click through the OAuth prompt and
    enter your Blockworks API key when asked.
  </Tab>

  <Tab title="ChatGPT">
    ChatGPT reaches remote MCP servers through developer-mode connectors.

    <Steps>
      <Step title="Open connector settings">
        Go to [chatgpt.com/admin/ca#available](https://chatgpt.com/admin/ca#available).
      </Step>

      <Step title="Enable developer mode">
        Turn on developer mode so custom connectors can be created.
      </Step>

      <Step title="Create the connector">
        Click **Create**.
      </Step>

      <Step title="Enter the server URL">
        Use `https://mcp.blockworks.com/mcp`.
      </Step>

      <Step title="Leave authentication as OAuth">
        Do not switch to another authentication type, because the server uses OAuth.
      </Step>

      <Step title="Create and authenticate">
        Click **Create**, then enter your Blockworks API key when prompted.
      </Step>
    </Steps>

    <Note>
      Developer-mode connectors may be restricted by your workspace administrator. If the option is missing, ask
      your ChatGPT workspace admin to enable it.
    </Note>
  </Tab>

  <Tab title="Other clients">
    Any MCP client that supports remote servers can connect with:

    * **URL**: `https://mcp.blockworks.com/mcp`
    * **Transport**: HTTP
    * **Authentication**: OAuth 2.0, handled by the client

    Most clients use the same config shape:

    ```json theme={null}
    {
      "mcpServers": {
        "blockworks": {
          "type": "http",
          "url": "https://mcp.blockworks.com/mcp"
        }
      }
    }
    ```

    Some clients omit `type` and infer the transport from the URL. If yours rejects the field, drop it and keep
    `url`.

    <Warning>
      Clients that only support local `stdio` servers cannot connect to a hosted URL directly. You need a client
      with remote server and OAuth 2.0 support.
    </Warning>
  </Tab>
</Tabs>

## Verify the connection

Whichever client you used, confirm the tools actually loaded before you rely on an answer:

```text theme={null}
What Blockworks tools do you have available?
```

If the assistant lists Blockworks tools, you are connected. If it lists none, or answers from general knowledge
instead of calling a tool, work through [Troubleshooting](/mcp/troubleshooting).