Apify Public Data MCP is a toolkit providing 40 production-grade public data extractors, targeted at MCP (Model Context Protocol) and the Apify platform. Each extractor runs as an Apify Actor on the user's own Apify account, performs read-only operations only, and does not collect data.
In terms of use, the toolkit covers four major categories of public data tasks: business intelligence and lead generation (such as Google Maps business leads, cross-border company registration queries), recruiting and market consumer intelligence (such as LinkedIn and Glassdoor jobs, Google News, Google Autocomplete), government and public records (such as SEC EDGAR, USAspending, Grants.gov, TED tenders, NHTSA recalls, FEC, EPA, state business registrations, contractor licenses), and research and health data (such as ClinicalTrials.gov, openFDA, Europe PMC, GLEIF, CMS providers, Census geocoding).
In terms of main capabilities, each Actor has built-in strict schema validation, deterministic field mapping, and self-healing DOM selectors, and uses event-based billing (starting at $0.00015 per result, starting at $0.0005 per Actor start; actual prices are subject to the live pricing on the Apify Store). Since version 1.1.0, calls return a {results, status, run?} structure, where status can be success, partial, empty_unverified, or error; a zero-row result does not necessarily mean there are no matching records. By default, each call requests a 120-second Actor timeout and a $1 event billing cap, adjustable via APIFY_MAX_CHARGE_USD; at most four calls run concurrently. Version 1.2.0 adds the APIFY_TOOL_PROFILE environment variable, which can be all (default), leads, market, government, or research, used to limit tool discovery and invocation scope.
In terms of usage, it can be installed to Claude Desktop, Cursor, or any MCP client via uvx apify-data-scrapers; it also supports Docker container operation or a local Python Stdio runner; Python and Node.js can call the Actor API directly. No-code users can use prebuilt blueprints for n8n, Make, Zapier, or Google Sheets; the repository includes eight n8n workflows and eight Make templates, as well as preconfigured tasks on the Apify Store.
Applicable scenarios include: AI agent integration (called via MCP), automated lead list building, SEC filing monitoring, job posting aggregation, government procurement opportunity alerts, competitor news monitoring, vehicle recall monitoring, and more. The repository also provides eight verified sample datasets, hosted on Hugging Face Datasets, which can be used for benchmarking or model training.
In terms of limitations: data from public sources may be delayed, incomplete, or temporarily unavailable; the MCP server runs locally, reads APIFY_TOKEN from environment variables, and only makes outbound calls to the Apify API; the toolkit does not provide a standalone subscription, and Actor runs are billed at list price through the user's Apify account; platform fees may apply. Private queued Actors are not included in this public catalog. The Node.js files are only Actor API examples, not a second MCP server. The author is Joseph McRell, under the MIT license.
