Web Data for AI Agents

Give an agent a reliable way to fetch structured web data at run time — every Actor is callable over a documented HTTP API, takes a schema-validated input and returns a schema-defined dataset.

How it works

These Actors are designed to chain — the output of one is the input of the next.

  1. 1

    Pick the Actor as a tool

    Each Actor’s input schema doubles as a tool definition: named parameters, types and required flags are already declared.

  2. 2

    Call it synchronously

    The run-sync-get-dataset-items endpoint runs the Actor and returns results in one request, which suits an agent tool call.

  3. 3

    Consume typed output

    The dataset schema tells the agent what fields to expect, so parsing is deterministic rather than guessed from prose.

Scrapers for this use case

5 Goldmine Actors serve this workflow.

What this does not solve

Worth knowing before you build a process on top of it.

  • Runs are asynchronous work with real latency. Synchronous calls are suitable for small inputs, not for large crawls inside a single agent turn.
  • Goldmine does not currently publish a hosted MCP server. Actors are reached over the Apify API, and Apify’s own MCP integration is the route for MCP-aware clients.
  • Costs scale with results, so an agent given an unbounded input can spend real money — bound it with `maxItems`.

Other use cases