{
  "markdown": "# StackScan MCP server\n\nAsk your assistant what a website runs, and who is behind it.\n\nA [Model Context Protocol](https://modelcontextprotocol.io) server for the\n[StackScan](https://www.stackscan.com) Tech Lookup API. Point Claude, Cursor, or any\nother MCP client at it and you can ask things like:\n\n> What is example.com built on?\n>\n> Who's behind vercel.com? Where are they, what industry, what's their LinkedIn?\n>\n> How many sites run Klaviyo, and in which countries?\n\nRead-only by design. Nothing here writes to any system of yours.\n\n---\n\n## Install\n\nYou need a StackScan account. Sign in, open **API Tokens** in the dashboard, create a\ntoken, and copy your workspace UUID from the same page.\n\n### Claude Desktop (extension)\n\nDownload `stackscan.mcpb` from the\n[releases page](https://github.com/stackscan/stackscan-mcp/releases), then drag it onto\n**Settings > Extensions**. It asks for your API token and workspace ID and installs the\nrest itself.\n\n### Claude Desktop (manual config)\n\n**Settings → Developer → Edit Config**, then add:\n\n```json\n{\n  \"mcpServers\": {\n    \"stackscan\": {\n      \"command\": \"npx\",\n      \"args\": [\"-y\", \"@stackscan/mcp-server\"],\n      \"env\": {\n        \"STACKSCAN_API_TOKEN\": \"your-token\",\n        \"STACKSCAN_TENANT_ID\": \"your-workspace-uuid\"\n      }\n    }\n  }\n}\n```\n\nRestart Claude Desktop. The tools appear under the connectors icon.\n\n### Claude Code\n\n```bash\nclaude mcp add stackscan \\\n  --env STACKSCAN_API_TOKEN=your-token \\\n  --env STACKSCAN_TENANT_ID=your-workspace-uuid \\\n  -- npx -y @stackscan/mcp-server\n```\n\n### Cursor, Windsurf, and others\n\nSame shape: `command: npx`, `args: [\"-y\", \"@stackscan/mcp-server\"]`, and the two\nenvironment variables. Check your client's own MCP docs for where its config file lives.\n\n---\n\n## Tools\n\n| Tool | What it does | Cost |\n|---|---|---|\n| `check_credits` | Remaining balance and this month's allocation | free |\n| `lookup_company` | The company behind a domain: name, industry, city, country, LinkedIn | 1 credit |\n| `lookup_domain_technologies` | Technologies on ONE domain, in full: category and global usage for each (`limit`, default 50, max 100) | 1 credit |\n| `lookup_technology` | How many sites run a technology, and where they are | 1 credit |\n| `lookup_companies` | The companies behind up to 20 domains, as a compact table | 1 credit per domain with data |\n| `lookup_domains_technologies` | The technologies on up to 20 domains, as a compact table. Optional `category` filter | 1 credit per domain with data |\n\nA lookup that finds nothing is **not** charged. Neither is `check_credits`.\n\n### Breadth vs depth\n\nThere are two technology tools and they answer different questions.\n`lookup_domains_technologies` is the **breadth** view: many domains, technology\nnames only, ideal for \"which of these run Shopify?\". `lookup_domain_technologies`\nis the **depth** view: one domain, every technology with its category and how many\nsites use it globally. The batch tool reports how many technologies it left out per\ndomain, so the model can offer to drill in.\n\n### Why the batch tools stop at 20\n\nThe REST endpoints behind them take 100 domains per request, and these tools\ndeliberately do not. A tool result goes straight into the model's context, and a\nhundred full company payloads is tens of thousands of tokens, which crowds out the\nconversation you are actually having, and the model then has to re-read all of it\nto answer anything. Twenty compact rows is a table a model can reason over.\n\nIt also refuses rather than truncates. If a batch would cost more than your session\ncap allows, it charges nothing and tells you how many you can afford. Quietly\ndropping domains would hand back an answer that looks complete and is not.\n\nIf you genuinely need hundreds of domains, that is what the\n[REST endpoint](https://api.stackscan.com/docs) is for.\n\n---\n\n## Configuration\n\n| Variable | Required | Default | Notes |\n|---|---|---|---|\n| `STACKSCAN_API_TOKEN` | yes | none | Dashboard → API Tokens |\n| `STACKSCAN_TENANT_ID` | yes | none | Your workspace UUID, same page |\n| `STACKSCAN_API_BASE` | no | `https://api.stackscan.com` | Rarely needed |\n| `STACKSCAN_SESSION_LOOKUP_CAP` | no | `25` | See below |\n\n### The session cap is a spend cap, and it is deliberate\n\nStackScan's own limit is a **rate** limit: requests per minute. That stops a burst. It\ndoes nothing to stop a patient agent quietly spending an entire credit balance over an\nafternoon, which is a failure mode humans clicking buttons do not have.\n\nSo this server keeps its own counter: **25 credit-consuming lookups per session** by\ndefault, after which it refuses and says why. Misses and `check_credits` never count\nagainst it, because they are never charged. Every successful result carries a footer\nshowing lookups remaining and your last known balance, so the model can pace itself\ninstead of discovering the limit by hitting it.\n\nIf you mean to go higher, raise it deliberately (`\"STACKSCAN_SESSION_LOOKUP_CAP\": \"200\"`) rather\nthan finding out afterwards.\n\n---\n\n## Development\n\n```bash\nnpm install\nnpm run build\n```\n\n`test-stdio.mjs` is a minimal JSON-RPC client that drives the server over a real stdio\npipe, so you can exercise it without an MCP host:\n\n```bash\nexport STACKSCAN_API_TOKEN=...\nexport STACKSCAN_TENANT_ID=...\nnode test-stdio.mjs check_credits lookup_company:stripe.com\n```\n\n**Never write to stdout in this server.** stdout is the JSON-RPC channel; anything else\non it corrupts the protocol. Diagnostics go to stderr.\n\n---\n\n## Notes for anyone reading the source\n\nThree things about the upstream API shaped this code and are easy to get wrong:\n\n**A miss is HTTP 200, not 404.** The API answers \"no data for that domain\" with\n`200 {\"success\": false, \"error\": \"No data available\"}`. A client that trusts the status\ncode reads the absent payload and crashes. Every response goes through one place that\nseparates a miss from a failure, so only a real failure is reported as an error.\n\n**Cached responses are still charged.** The API's response cache shields its database,\nnot your wallet: a repeat lookup inside the cache window costs the same credit. That is\nwhy the local balance decrements on every hit instead of trying to guess which calls\nwere free.\n\n**Two base URLs are live.** The API's home is `api.stackscan.com/v1/...`. The older\n`app.stackscan.com/api/v1/...` paths are still served and are deliberately *not*\nredirected, because a redirect would drop the body on POST. If your config still names\nthe old host, this server detects it and keeps using the old path shape, so upgrading\nwill not break you.\n\n## Privacy Policy\n\nStackScan's privacy policy is at\n[www.stackscan.com/privacy](https://www.stackscan.com/privacy). What it means for this\nserver specifically:\n\n**What is collected.** The domains and technology names you look up, sent to\n`api.stackscan.com` over HTTPS along with your API token and workspace ID. Nothing else\nleaves your machine. This server reads no files, opens no ports and keeps no local\nstore; the only state it holds is a lookup counter and a credit balance, both in memory\nand both gone when the process exits.\n\n**How it is used and stored.** Lookups are logged against your workspace so credits can\nbe metered and the account can be supported. Results come from StackScan's own crawl\ndata, not from anything you send.\n\n**Third parties.** Query terms are not sold or shared for advertising. They reach the\ninfrastructure providers StackScan runs on, as set out under \"Sharing With Service\nProviders\" in the policy, and nowhere else.\n\n**Retention.** Covered under \"Data Retention\" in the policy. Delete your account and\nthe associated lookup history goes with it.\n\n**Contact.** hello@stackscan.com, or [www.stackscan.com/contact](https://www.stackscan.com/contact).\n\n## Links\n\n- [API reference](https://api.stackscan.com/docs)\n- [What this is, in plain English](https://www.stackscan.com/mcp)\n- [StackScan](https://www.stackscan.com)\n\n## License\n\nMIT. See [LICENSE](LICENSE).\n\nThe licence covers this client only, roughly 500 lines that make HTTP requests.\nThe StackScan data it reaches is a paid service and is not covered by it.\n",
  "bytes": 8130,
  "sha": "04fe25f1a32935f738abcb7e6ae5d4ac82af295a2d953d3bcdbeea1714017291",
  "repo_slug": "stackscan/stackscan-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_com_stackscan_mcp_server_32b5a770/readme"
}