Using MCP

Run one command in your terminal. The second form (--scope user) registers the server for every project. Check the connection with claude mcp list.

Terminal

claude mcp add --transport http zenith-code https://mcp.zenithcode.net/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

claude mcp add --transport http --scope user zenith-code https://mcp.zenithcode.net/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

claude mcp list

To share the setup with a team, put the generic JSON config below in the project’s .mcp.json and pass the key through the ZENITH_API_KEY environment variable, so the key never lands in the repository.

From the app’s Connectors screen, connect with the steps in “Connect with OAuth” above. To use an API key in the config file instead, use the open-source mcp-remote bridge (requires Node.js), which forwards requests with the header attached.

claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\)

{
  "mcpServers": {
    "zenith-code": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.zenithcode.net/mcp",
        "--header",
        "x-api-key:${ZENITH_API_KEY}"
      ],
      "env": {
        "ZENITH_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

The header is written without a space after x-api-key: to avoid argument-splitting problems on Windows. Restart Claude Desktop after saving.

Add it to Cursor’s MCP config file.

~/.cursor/mcp.json (all projects) or .cursor/mcp.json (per project)

{
  "mcpServers": {
    "zenith-code": {
      "url": "https://mcp.zenithcode.net/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

After saving, check that the server is enabled in the MCP section of Cursor’s settings.

Add it to .vscode/mcp.json in your workspace to use the tools in Copilot Chat agent mode. With inputs, VS Code prompts for the key the first time and does not store it in the file.

.vscode/mcp.json

{
  "inputs": [
    {
      "type": "promptString",
      "id": "zenith-api-key",
      "description": "ZENITH Code API key",
      "password": true
    }
  ],
  "servers": {
    "zenith-code": {
      "type": "http",
      "url": "https://mcp.zenithcode.net/mcp",
      "headers": {
        "Authorization": "Bearer ${input:zenith-api-key}"
      }
    }
  }
}

Switch Copilot Chat to agent mode; if zenith-code appears in the tool list, you are connected.

In the ChatGPT app, connect with the steps in “Connect with OAuth” above. With the OpenAI Responses API, you can pass headers to an MCP tool and use an API key.

Python (openai package)

from openai import OpenAI

client = OpenAI()
resp = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "server_label": "zenith_code",
        "server_url": "https://mcp.zenithcode.net/mcp",
        "headers": {"Authorization": "Bearer YOUR_API_KEY"},
        "require_approval": "never",
    }],
    input="How many years in a row has Toyota (7203) raised its dividend?",
)
print(resp.output_text)

Setting require_approval to never skips the confirmation for each tool call. Replace the model name with the one you use.

Most clients that support Streamable HTTP and headers accept this shape (it is also the format of Claude Code’s .mcp.json).

.mcp.json or similar (file and key names vary by client)

{
  "mcpServers": {
    "zenith-code": {
      "type": "http",
      "url": "https://mcp.zenithcode.net/mcp",
      "headers": {
        "Authorization": "Bearer ${ZENITH_API_KEY}"
      }
    }
  }
}

The part after Bearer reads the key from the ZENITH_API_KEY environment variable. If your client does not expand environment variables, write the key directly. Some clients spell the type as streamable-http.