queryfanout.com
Account

Settings

Model API keys

Runs call Gemini and OpenAI with your own keys. They're saved in this browser only and sent with each run; the server never stores them.

API tokens

Tokens let the MCP server, sales-automation tools and the API act as you. Treat them like passwords. The Chrome extension shares anonymously without one; add a token there only if you want its captures under My runs.

Connect queryfanout to Claude

The MCP server lets Claude run fan-outs, search the index and read your runs from a conversation. Model keys travel in request headers, so they never enter the model's context, and live runs are billed to your own keys exactly as they are here.

Endpointhttps://qfo-api-52830742425.us-central1.run.app/mcp/
run_fanoutRuns a live fan-out on Gemini and/or OpenAI, with optional real-search passes· needs model keys
search_indexSearches every fan-out query in the open index, by keyword and meaning
reverse_fanoutFinds the prompts that fan out to a query, page topic or keyword
get_runLoads a saved run from its id or queryfanout.com/r/… link
list_my_runsLists your own runs and Chrome extension captures

Replace qfo_YOUR_TOKEN with a token from API tokens above (sign in first); create one and it's filled in for you.

  1. 1
    Create an API token
    Use API tokens above. Name it something like “Claude Desktop” so you can revoke it on its own later.
  2. 2
    Check that Node.js 18 or newer is installed
    Claude Desktop connects to remote servers like this one through mcp-remote, which runs with Node. In a terminal, run node --version; if it's missing or older than 18, install the LTS version from nodejs.org.
  3. 3
    Open Claude Desktop's config file
    From the Claude menu in your system menu bar (not the settings inside the chat window), choose Settings…, open the Developer tab and click Edit Config. The file is %APPDATA%\Claude\claude_desktop_config.json on Windows and ~/Library/Application Support/Claude/claude_desktop_config.json on macOS.
  4. 4
    Add queryfanout to the file and save it
    If the file is empty, paste this in. If it already has an mcpServers section, add the queryfanout entry inside it. The header values live in env with no spaces in args, which avoids a Windows bug with spaces in arguments.
    claude_desktop_config.json
    {
  "mcpServers": {
    "queryfanout": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://qfo-api-52830742425.us-central1.run.app/mcp/",
        "--header",
        "Authorization:${QFO_AUTH}",
        "--header",
        "X-Gemini-Key:${GEMINI_KEY}",
        "--header",
        "X-OpenAI-Key:${OPENAI_KEY}"
      ],
      "env": {
        "QFO_AUTH": "Bearer qfo_YOUR_TOKEN",
        "GEMINI_KEY": "YOUR_GEMINI_KEY",
        "OPENAI_KEY": "YOUR_OPENAI_KEY"
      }
    }
  }
}
  5. 5
    Quit Claude Desktop completely and reopen it
    Then click the + button at the bottom left of the message box, point to Connectors and choose Manage connectors. You should see queryfanout with its five tools. If it's missing, check mcp.log in %APPDATA%\Claude\logs (Windows) or ~/Library/Logs/Claude (macOS).
  6. 6
    Try a prompt
    Pick one of the prompts below. Claude asks before each tool call; choose Always allow once you're comfortable.

Prompts to try

Mention queryfanout by name the first time so Claude reaches for the right tool. Live runs take 20 to 90 seconds.

run_fanout
Find the stable sub-queries

Use queryfanout to run a fan-out for "best project management software for agencies" on Gemini and OpenAI with 5 passes and 1 real-search pass. Which sub-queries did both models generate, and which ones came back in every pass?

run_fanout
Turn a fan-out into a content plan

Run a queryfanout fan-out for "how to choose a 401k provider for a small business". Group the sub-queries into content hubs, and for each hub suggest the page format to build based on each sub-query's routing format and query class.

search_index
Research a topic in the index

Search the queryfanout index for "ev winter range". What sub-queries have AI models generated around this topic, and which prompts did they come from?

reverse_fanout
Which prompts lead to my page?

Use queryfanout's reverse fan-out to find the prompts that fan out to "rivian r1s towing capacity". Which broader questions would retrieve a page about it, and how often does it come up?

get_run
Explain a run

Get the queryfanout run at https://queryfanout.com/r/RUN_ID and summarize where Gemini and OpenAI disagree, and which real searches the simulation predicted.

list_my_runs
Compare real and simulated searches

List my recent queryfanout runs and extension captures. For prompts that have both, compare the searches ChatGPT really ran with the simulated sub-queries and tell me what the simulation missed.

reverse_fanout + run_fanout
Find gaps on a page

Read https://example.com/guide. Use queryfanout's reverse fan-out on its main topic to find the prompts that lead to it, run a fan-out for the top prompt, and list the sub-queries this page doesn't answer.

run_fanout
Prep for a sales call

Run a queryfanout fan-out for the question a buyer of [prospect's product category] would ask an AI assistant. Draft three talking points for a sales call about the sub-queries [prospect's site] doesn't cover.