The best MCP server for AI search. The answers buyers see, in your coding agent.

    Add one server to Claude Code, Cursor or Codex. Your agent asks twelve AI engines your buyers' questions, reads every answer and checks Google rankings, on your Honeyb key at the API's prices.

    Set it up

    We issue keys within one working day. Tell us your engines and volume.

    Tools for your agent
    8
    AI engines, real browser sessions
    12
    Per 1,000 answers, from
    $1.05
    Claude Codehoneyb over MCP

    $ claude mcp add --transport http honeyb https://platform.honeyb.ai/mcp --scope project --header 'Authorization: Bearer ${HONEYB_API_KEY}'

    # writes the honeyb server into .mcp.json; the key stays in HONEYB_API_KEY

    $ claude

    > Ask ChatGPT, Google AI Overviews and Google AI Mode "best CRM for small businesses" for HubSpot in the US.

    ● honeyb · list_engines_and_prices

    12 engines, each a real browser session. $2.05 per 1,000 enriched answers on chatgpt, google_ai_overviews, google_ai_mode.

    That is 3 enriched answers, $0.0061 in credits. Run it?

    > yes

    ● honeyb · create_visibility_task

    3 answers held, $0.0061

    ● honeyb · get_task

    status: completed, 3 of 3 answers

    ● honeyb · get_task_results and get_task_citations

    Engine Names HubSpot First source Sources

    ChatGPT yes spotsaas.com 12

    Google AI Overviews yes reddit.com 4

    Google AI Mode yes reddit.com 9

    Setup

    One line in Claude Code. One click in Cursor.

    Every config reads the key from an environment variable, so it never lands in a file, a commit or the chat.

    Claude CodeCursorCodexVS Code
    Claude Code
    claude mcp add --transport http honeyb https://platform.honeyb.ai/mcp --scope project --header 'Authorization: Bearer ${HONEYB_API_KEY}'
    Codex, ~/.codex/config.toml
    [mcp_servers.honeyb]url = "https://platform.honeyb.ai/mcp"bearer_token_env_var = "HONEYB_API_KEY"

    Any agent: paste the setup prompt. It adds the server, reads your key from the environment and runs a first task.

    Claude Code. Run the one line. It adds the server to the project’s .mcp.json for the whole team.

    Cursor. Add to Cursor writes the same config in one click.

    Any agent. Paste the setup prompt. The agent adds the server, lists the engines and runs a first task after you confirm the price.

    Opening the server address in a browser brings you back to this page. Your agent connects to it with your key.

    Tools

    Eight tools. Two spend credits.

    Your agent quotes the price before it spends anything, then asks, polls and reads. Every tool runs the same operation as the REST API.

    • list_engines_and_prices

      Lists the engines a task can run on, whether each searches the web, and every operation with its price and limits.

      Free
    • create_visibility_task

      Asks the engines you pick a set of prompts about a brand and reads every answer: whether the brand is named, the competitors beside it, the sentiment and the pages cited.

      Spends credits
    • get_task

      Returns a task's status, its answer counts and the credits held and charged. The agent polls it until the task finishes.

      Free
    • get_task_results

      Returns every answer, one row per prompt and engine: the answer text, the brand and competitor mentions and the sentiment.

      Free
    • get_task_citations

      Returns every page the engines cited, with its domain, page classification, position and whether the citing answer named the brand.

      Free
    • create_rank_check

      Checks where a domain ranks on Google for up to 1,000 keywords: position, ranking URL, SERP features and the AI Overview citation.

      Spends credits
    • get_rank_check

      Returns a rank check's status and keyword counts, and collects the keywords Google has finished. The agent polls it about once a minute.

      Free
    • get_rank_check_results

      Returns one row per keyword: position, ranking URL and title, SERP features and whether the AI Overview cited the domain.

      Free

    Price

    The API's prices, over MCP.

    Tool calls spend the same prepaid credits as the REST API. Credits never expire, and a failed answer is never charged.

    Per 1,000 raw answers
    $1.05
    Per 1,000 enriched answers
    $2.05
    Per 1,000 Google rank checks
    $1
    Listing, polling and reading
    Free

    See every price·Why the app’s answer, not the model API’s

    Coverage

    Every engine your agent can ask.

    AI answers: run each prompt from any country, and ChatGPT prompts from a US state too. Google rank checks: any Google location code, from a country down to a city.

    ChatGPTReal browser sessionGoogle AI OverviewsReal browser sessionGoogle AI ModeReal browser sessionGeminiReal browser sessionBing CopilotReal browser sessionNaver AI BriefingReal browser sessionNaver AI ModeReal browser sessionERNIEReal browser sessionPerplexityReal browser sessionClaudeReal browser sessionGrokReal browser sessionDeepSeekReal browser session

    FAQ

    MCP server questions, answered.

    Honeyb API

    Give your agent the answers your buyers see.

    Twelve AI engines captured in real browser sessions, and Google rankings, as tools on one key.

    ChatGPTClaudeGeminiPerplexity
    See pricing

    We issue keys within one working day. Tell us your engines and volume.