The MCP server is here

Give your coding agent or AI assistant live answers from AI search: it asks ChatGPT, Gemini, Perplexity and the other engines and reads the results on its own.

What you get

• run_prompt: sends a prompt to any engine and waits for the answer, in one tool call.
• submit_prompt and get_task: queue a longer job and read its result later.
• The same API key, engines and credit prices as the REST API.

How it works

MCP is the open standard AI agents use to call outside tools. Add the querying.ai server to your agent’s MCP settings with your API key, and the agent asks AI search engines questions while it works: it researches a market, checks how engines describe a brand or gathers sources for a draft.

The agent chooses the engine and country for each call and receives the same JSON your code gets from the API, so its findings rest on real answers.

Where it helps

• Ask your coding agent how ChatGPT and Gemini answer a question before you write content for it.
• Let a research agent collect answers and sources across engines and summarize them for you.

Get started

Add the server where your agent keeps its MCP servers. Claude Code and Codex read the key from the QUERYING_API_KEY environment variable, so export it in the shell that starts the agent.

claude mcp add --transport http querying \
  https://mcp.querying.ai/mcp \
  --header 'Authorization: Bearer ${QUERYING_API_KEY}'

For any other client, paste the mcp.json entry and replace YOUR_API_KEY with a key from the dashboard. The server speaks Streamable HTTP at https://mcp.querying.ai/mcp.

Tools

Once connected, the agent calls these tools. Monitoring tools read and write the same monitors as your dashboard, and every task appears in Activity like an API call.

run_prompt

Sends a prompt and waits for the answer, up to the tool’s own timeout: 180 seconds by default, 900 at most. A task still running then comes back as a task ID for get_task.

submit_prompt

Sends a prompt and returns the queued task at once.

get_task

Reads the status, or the finished response, of a task you sent.

research_prompts

Takes a topic and returns the questions people ask AI assistants about it, ranked by monthly search demand, with the set a monitor should track. KR and US, 12 credits.

scrape_page

Takes one web page and returns its main content as Markdown, with images, title, author and publish date. 1 credit per page.

create_monitor

Creates a scheduled monitor. Each run spends the credits of prompts × engines tasks, on the monitor’s schedule.

run_monitor

Starts a run now, for the credits of one scheduled run.

get_monitor_analytics

Rates, the previous period of equal length, changes in percentage points, rates by engine, share of voice, opportunities and coverage.

get_monitor_results

The scored answer rows. includeEvidence adds the stored answer text and citations.

get_monitor_answer

One answer in full: Markdown, sources, and the aliases and competitors it was scored against.

Pricing

Tasks started through MCP use the same credits as the API and are charged when they complete.