Hubertino: Google Maps leads

Hubertino MCP server: Google Maps leads for Claude, Cursor and other AI agents

@hubertino/mcp-server connects any Model Context Protocol client (Claude Desktop, Claude Code, Cursor, VS Code, ...) to Hubertino, a Google Maps lead scraper. Ask your assistant for "dentists in Austin with emails". It starts a scrape, waits for it to finish, and gives you the leads as a table or as a CSV/XLSX file.

It wraps Hubertino's authenticated public REST API (docs) and uses your own API key and credits.

What you get

Each lead is one Google Maps business. Only fields that have a value are returned:

• Contact: name, category, phone, website, and an email found on the business's own website (enrichWebsite).
• Social profiles: Facebook, Instagram, LinkedIn, X, YouTube (enrichLinks).
• Location: address, city, state, postal code, country code, and a Google Maps link / place id.
• Reputation: rating, review count, business status.
• Extra fields (left out of compact results to save context): coordinates, opening hours, price range, service options, and order/reservation/menu links. Get them with get_results detail: "full", or in the download_export file.

Tools

There is also one prompt, build_lead_list, which runs the whole workflow for a business type and area.

start_scrape takes the same inputs and limits as the API:

One scrape runs at most 2,000 searches (categories.length × locations.length).

Requirements

• Node.js 20 or newer (npx must be on the PATH your MCP client uses).
• A Hubertino account and an API key. New accounts get 100 free credits, and the API is included on every plan, including the free one.

1. Sign up at hubertino.com.
2. Open Settings (/dashboard/settings) and create an API key. It starts with hub_live_ and is shown only once, so copy it straight away.

Install

Every client needs the same thing: run npx -y @hubertino/mcp-server with HUBERTINO_API_KEY set in its environment. Replace hub_live_your_key_here with your key.

Claude Desktop

Open Settings → Developer → Edit Config. This opens claude_desktop_config.json, which lives at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS and %APPDATA%\Claude\claude_desktop_config.json on Windows. Add:

{

"mcpServers": { "hubertino": { "command": "npx", "args": ["-y", "@hubertino/mcp-server"], "env": { "HUBERTINO_API_KEY": "hub_live_your_key_here" } } } }

Restart Claude Desktop. The Hubertino tools then show up in the tools menu.

A one-click bundle (hubertino-mcp.mcpb) can be built with npm run bundle and opened in Claude Desktop. It asks for the key in a form, so you don't edit any JSON.

Claude Code

claude mcp add hubertino -e HUBERTINO_API_KEY=hub_live_your_key_here -- npx -y @hubertino/mcp-server

Add --scope user to make it available in every project. Run claude mcp list to check that it's connected.

Cursor

Add the server to ~/.cursor/mcp.json (all projects) or to .cursor/mcp.json (one project):

{

"mcpServers": { "hubertino": { "command": "npx", "args": ["-y", "@hubertino/mcp-server"], "env": { "HUBERTINO_API_KEY": "hub_live_your_key_here" } } } }

VS Code (GitHub Copilot agent mode)

Add this to .vscode/mcp.json. VS Code prompts for the key and stores it securely:

{

"inputs": [ { "type": "promptString", "id": "hubertino-key", "description": "Hubertino API key (hub_live_...)", "password": true } ], "servers": { "hubertino": { "type": "stdio", "command": "npx", "args": ["-y", "@hubertino/mcp-server"], "env": { "HUBERTINO_API_KEY": "${input:hubertino-key}" } } } }

Other clients / from source

Any MCP client that can launch a stdio server works. To run from a checkout:

git clone https://github.com/hubertino-app/hubertino-mcp.git

cd hubertino-mcp && npm install && npm run build

then point your client at: node /absolute/path/to/hubertino-mcp/dist/index.js

Configuration

Example prompts

• "Find dentists in Austin, TX with emails. 100 per search is enough."
• "Build a list of plumbers and electricians in Leeds and York, UK, then save it as a CSV."
• "Scrape roofing contractors in these zip codes: 78701, 78702, 78703. Show only the ones with an email and a website."
• "How is my last scrape doing?"
• "Show me page 2 of the results from my cafe scrape: name, phone and Instagram only."
• "Which of my finished scrapes had the most results? Download it as Excel."

A typical run looks like this:

start_scrape     {"categories":["dentist"],"locations":["Austin, TX, United States"],"maxResults":100}

-> {"id":"6f1c…","status":"queued","credits":"Reserved 100 credits … the rest is refunded automatically.", …} wait_for_scrape {"scrapeId":"6f1c…"} -> {"status":"done","resultCount":…,"creditsCharged":…,"coverage":{"total":…,"email":…}, …} get_results {"scrapeId":"6f1c…","withEmailOnly":true,"limit":50} -> {"total":…,"returned":…,"hasMore":true,"nextOffset":50,"rows":[{"name":"…","phone":"…","email":"…"}, …]} download_export {"scrapeId":"6f1c…","format":"csv"} -> {"path":"/Users/you/Downloads/hubertino-6f1c….csv","bytes":…}

Credits, limits and errors

• 1 credit = 1 business delivered. When a scrape starts, Hubertino reserves categories × locations × maxResults credits, capped at your balance. When it finishes, you are charged only for rows actually delivered and the rest is refunded. If your balance is below the worst case, the scrape still runs but stops once the balance is used up. See pricing.
• Speed: a scrape takes minutes. Rows can be read while it runs (partial: true). Each row's _status goes found → extracted → enriched.
• Rate limit: 12 new scrapes per 5 minutes per account. Status, results and export calls are not rate limited.
• Page size: get_results reads up to 1,000 rows per call (default 50). To protect the model's context window, a page that would not fit in one tool response (about 50,000 characters) is cut short with trimmed: true; keep paging from nextOffset, ask for fewer columns with fields, or use download_export to get everything.
• Long scrapes: get_scrape and wait_for_scrape echo at most 25 categories/locations and add categoriesTotal / locationsTotal for longer lists. wait_for_scrape never runs more than about 8 s past its timeoutSeconds, even when the API answers slowly.
• Export: only available once status is done.

Errors come back as tool errors with a next step, so the assistant can recover or tell you what to do:

If starting a scrape fails with a timeout, network error or 5xx, the scrape may still have been created. start_scrape then says to check list_scrapes before starting it again, so the same scrape is not run and charged twice.

Security and privacy

• Your API key stays in your MCP client's config and is sent only as a Bearer header to the Hubertino API over HTTPS. The server never logs it or returns it in a response. Don't paste your key into the chat.
• The server calls only the authenticated /api/v1 endpoints listed above. It makes no other network calls and does no scraping itself.
• download_export writes only hubertino- .csv / .xlsx files into the chosen folder and never overwrites an existing file.
• Use lead data lawfully. Follow the privacy and anti-spam rules that apply to you and to the people you contact (for example GDPR, PECR, CAN-SPAM).

Development

npm install

npm run build # compile to dist/ npm test # type-check and run the node:test suites (mocked fetch; no network) npm run smoke # spawn dist/index.js over stdio and list tools (set HUBERTINO_API_KEY for a live read-only call) npm run bundle # build hubertino-mcp.mcpb (Claude Desktop / Smithery bundle)

optional: compare mirrored limits with a local checkout of the Hubertino app

HUBERTINO_APP_DIR=/path/to/hubertino/app npm test

Publishing steps are in PUBLISHING.md.

Links

• API docs: https://hubertino.com/docs
• Pricing: https://hubertino.com/pricing
• Google Maps scraper: https://hubertino.com/google-maps-scraper
• Affiliate program (25% recurring): https://hubertino.com/partners

License

MIT © Hubertino MB