io.github.Mamoo85/open-data-actors-mcp

open-data-actors-mcp: MCP server for public-data Apify actors (sanctions, SEC Form D, WARN, FMCSA)

A thin Model Context Protocol server that exposes 29 public-data Apify actors as tools: sanctions screening, SEC Form D funding leads, WARN layoff notices, FMCSA new carriers, technical SEO audits, tech-stack detection, job boards, building permits, Google Maps places, TikTok profiles, YouTube comments, Amazon best sellers, SAM.gov federal bids, and more.

Disclosure: I am the author and maintainer of these actors. They run on Apify and are pay-per-result; the Apify free plan credit covers a trial. This server only forwards your tool call to the Apify API with your token. It has no backend of its own, stores nothing, and sends nothing anywhere else. If you want Apify's general-purpose MCP server (every Store actor), use mcp.apify.com; this one is a small, fixed, typed subset.

Hosted (no install)

Stateless streamable-HTTP endpoint: https://open-data-actors-mcp.vercel.app/mcp

Bring your own Apify token on every request (Authorization: Bearer <token>, or ?token=). Nothing is stored or logged by us; runs bill to your own Apify account. Rate limit: 30 requests/minute per token.

claude mcp add --transport http open-data-actors https://open-data-actors-mcp.vercel.app/mcp --header "Authorization: Bearer $APIFY_TOKEN"

Verify with a real client: APIFY_TOKEN=... node test/http.test.mjs.

Install

You need Node 18+ and an Apify API token (https://console.apify.com/settings/integrations?fpr=rnmakg).

{
  "mcpServers": {
    "open-data-actors": {
      "command": "npx",
      "args": ["-y", "github:Mamoo85/open-data-actors-mcp"],
      "env": { "APIFY_TOKEN": "your-apify-token" }
    }
  }
}

(Not on npm yet; the github: form installs straight from this repo.) Works in Claude Desktop, Claude Code (claude mcp add open-data-actors -e APIFY_TOKEN=... -- npx -y github:Mamoo85/open-data-actors-mcp), Cursor, and any stdio MCP client.

Tools

Each tool takes the actor's own input fields plus max_results (default 25, cap 200). Calls run synchronously (run-sync-get-dataset-items, 240 s timeout) and return {count, items}. Fields that are secrets or state (proxy settings, API keys, webhook URLs, incremental-state names) are intentionally not exposed.

Cost

Each call starts an Apify run and you pay Apify the actor's pay-per-result price for the rows it produces. Lower max_results to cap spend. See each Store page for current pricing.

Tests

npm test starts the server and drives it with the real MCP SDK client over stdio: lists tools, checks error handling, and (when APIFY_TOKEN is set) makes one live tools/call. Verified 2026-10-14: 29 tools listed, live sanctions_screening call returned data.

Development

python gen.py regenerates src/tools.json from the actors' published input schemas.

MIT licensed.

Some links are affiliate links.