Skip to main content
We provide an MCP server to use the indices API from your AI agents and applications like ChatGPT and Claude. This allows your agents to create and invoke tasks automatically.

​
Endpoint

MCP endpoint: https://mcp.indices.io The MCP server requires API key authentication. Your requests must contain a header Authorization: Bearer <your_api_key>. Instructions below on how to include this easily with agents like Claude.

​
Usage

First, generate an API key in the dashboard. We recommend using a separate API key for each agent. Then add the indices MCP server to your AI agent or application. Use the Streaming HTTP transport, with the URL https://mcp.indices.io, and ensure requests will contain the Authorization header with your API key.

​
Adding to Claude

​
Setup

With Claude, add the following MCP server to your configuration: 
{
  "mcpServers": {
    "indices-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.indices.io",
        "--header",
        "Authorization: Bearer ${AUTH_TOKEN}"
      ],
      "env": {
        "AUTH_TOKEN": "sk-<paste your api key here>"
      }
    }
  },
  ...
}
Behind the scenes, the mcp-remote library proxies the requests to our MCP server, adding the Authorization header to all requests to authenticate them. mcp-remote isn’t always necessary if your MCP client supports adding headers to outbound requests, as Claude Code does for example. If you’re using Claude Code, you can either first add the MCP server to Claude Desktop first and then run claude mcp add-from-claude-desktop, or you can run claude mcp add indices-mcp https://mcp.indices.io --transport http --header "Authorization: Bearer <paste your API key here>". This leads to a config (in ~/.claude.json) like: 
"indices-mcp": {
  "type": "http",
  "url": "https://mcp.indices.io",
  "headers": {
    "Authorization": "Bearer <your API key>"
  }
}
Note that mcp-remote was not needed here. Refer to the MCP documentation for further guidance on adding MCP servers to typical MCP clients, or the documentation for your specific MCP client. OpenAI’s Responses API supports passing through Authorization headers as well.

​
Interaction

Now, you can chat with Claude and ask it to perform tasks using the MCP server. In the following example, we already have a task set up to retrieve AI events in London from Luma. We can ask the agent Can you get a list of Luma events happening today (via indices.io MCP)? The agent responds, informed by live data from Luma, retrieved via the indices platform! 
Found 9 Luma events happening in London today (October 2, 2025):

  Morning:
  - Annual Town Hall (7:30 AM)
  The Trampery Old Street
  https://luma.com/event/d7ukgqri
  - AEO LDN October: Discover How To Rank in AI Searches (8:00 AM)
  Crypt on the Green, St James Church
  https://luma.com/event/7on6hba8

  Afternoon:
  - M&A Round Table: A data-led approach (12:30 PM)
  Malmaison London
  https://luma.com/event/1hthvio3
  - Me, Myself & AI: The Workshop on How to Use AI for Work and Play (4:00 PM)
  Bluebird City
  https://luma.com/event/z22rlt26
  - After Work: App Leaders London (4:30 PM)
  Maison Assouline, Piccadilly
  https://luma.com/event/aarki-after-work-app-leaders-london

  Evening:
  - Agentic Happy Hours after MCP Dev Summit (5:00 PM)
  London, England
  https://luma.com/event/MCP-Dev-Summit-Happy-Hours
  - AI Engineer London Meetup #7 (5:00 PM)
  London, England
  https://luma.com/event/ubg0u3a2
  - MCP Happy Hour by goose (5:30 PM)
  18 Old Bailey
  https://luma.com/event/0ipevcm6
  - Beyond vibe coding: когда ИИ начнёт писать весь код за нас? (6:00 PM)
  Rich Mix, Bethnal Green
  https://luma.com/event/qdqls5xz
Luma events happening today, retrieved via indices