n8n, Make and Zapier

Automation platforms can call the PublicWWW tools as steps of a workflow - run a search on a schedule, put the sites into a spreadsheet or a CRM, or let an AI step decide what to search for. They connect to the same server with a token.

Issue a token for each platform on your profile page, so that any of them can be cut off alone. Free accounts get the same tools with the free plan's limits - see Limits and errors. For a plain scheduled export without a tool step, the REST API is simpler: every platform has an HTTP step, and the API is described in OpenAPI.

n8n

Two nodes speak MCP. MCP Client runs one tool as an ordinary step; MCP Client Tool hands the tools to an AI Agent node, which picks and calls them itself. Both take the same settings:

In MCP Client choose the tool - search or account - and fill in its parameters, or switch Input Mode to JSON:

{ "query": "\"angular.min.js\"", "rows": 100 }

The node returns one item with the tool's answer in structuredContent: the sites in results (domain, url, rank), and total, returned, truncated next to them. To get one item per site, follow it with a Split Out node on structuredContent.results.

Make

Add the MCP Client app to a scenario. In its connection choose + New MCP Server, enter https://api.publicwww.com/mcp as the URL and paste your token into API Key / Access token. Then use Call a tool with search to run a search as a step, or Execute an action with AI to let the AI choose the calls.

Call a tool lists the tool's arguments - query, rows, page, snippets. Its output carries the answer twice: Result as JSON text and Parsed Result as fields - results with the sites, total, returned, truncated and the rest. To handle the sites one at a time, add an Iterator over Parsed Result > results.

Zapier

Connect the MCP Client app: server URL https://api.publicwww.com/mcp, transport Streamable HTTP, OAuth off, and your token in the bearer token field. Zapier reads the tool list right away - search and account with their parameters. For a search use the action Run Read-Only Tool with search; the trigger New Tool only reports tools added to the server.

Into a CRM: HubSpot, Pipedrive, Salesforce

The common case is a list of prospects: every site that loads a certain script, widget or tracking code, put into the CRM as companies. All three platforms above have steps for HubSpot, Pipedrive and Salesforce, so the workflow is the same everywhere:

1. A trigger - a schedule, or a button you press.
2. The PublicWWW search tool with your query, for example "googletagmanager.com/gtm.js" site:de, and as many rows as you want to import.
3. A loop over results: one site per pass, with its domain, url and rank.
4. Look the company up in the CRM by its domain, and create it only if it is not there - running the workflow again then adds the new sites and leaves the rest alone.

It helps to note on the record why it is there - the query in a custom field or a note - so a later search for something else does not mix the lists. The rank sorts the list by popularity: 1 is the most popular site, and an empty rank means the site has no popularity rank.

What to watch

• Each run of a search is one search from the account's daily allowance, and at most 10 requests a minute go through per account - a loop that calls the tool for every row of a sheet runs into both. Ask for more rows in one search instead.
• Refusals come back as tool errors with a code - quota_exceeded, too_many_requests with retry_after - which the platforms show as a failed step. The codes are in Limits and errors.