Skip to content

Cursor

Configure an API key from the Klozeo dashboard, as shown below, or use browser login.

  1. Add Klozeo to your project’s .cursor/mcp.json:

    {
    "mcpServers": {
    "klozeo": {
    "url": "https://mcp.klozeo.com/mcp",
    "headers": { "Authorization": "Bearer sk_live_your_key" }
    }
    }
    }

    Replace the placeholder with your key. Keep this file private and out of version control, or use your private user MCP configuration.

  2. Enable the server in Cursor’s MCP settings. For Cursor Agent CLI, approve this configured server with:

    Terminal window
    cursor-agent mcp enable klozeo
    cursor-agent mcp list-tools klozeo
  3. Check that tools such as list_leads and get_stats appear. In a chat, ask:

    List my five most recent leads in Paris.

Leave out headers:

{
"mcpServers": {
"klozeo": { "url": "https://mcp.klozeo.com/mcp" }
}
}

In Cursor’s MCP settings, click Connect (or Needs login) next to klozeo. Cursor opens Klozeo in your browser; sign in, select one or more workspaces and approve access.

Browser login gives the twelve assistant tools with read access first; writing needs a separate approval in Klozeo. It can reach several workspaces; an API key reaches only the workspace it was created in. With several workspaces, the assistant lists them with list_workspaces and names the workspace in each call (see Several workspaces). Keys, webhooks, scoring, attributes, deletions and batches need an API key. See Authentication.

HTTP 401 means the key is missing, invalid or revoked. Replace it and reconnect; a session established with another key cannot be reused. HTTP 503 means the API cannot currently validate your key; retry later.

The separate OpenAI endpoint /openai/v1/mcp does not accept API keys.

The local CLI recipe uses Cursor Agent 2026.09.18-9a7762b, an isolated project configuration and a synthetic API-key fixture. The desktop Cursor UI and LLM tool selection are separate checks; this recipe does not certify the deployed service.