@synoppy/mcp

Give your AI agent the whole web. An MCP server for Synoppy — drop it into Claude Desktop, Cursor, or any MCP client and your agent gets tools to read, search, crawl, map, extract, classify, enrich, screenshot, and pull images from the live web.

Get a free key → · Docs · synoppy.com

Every tool returns the full JSON response, including the metered billing fields creditsUsed and creditsRemaining.

Tools

| Tool | Endpoint | Inputs | | --- | --- | --- | | synoppy_read | POST /api/scrape | url, formats? (markdown|html|text), onlyMainContent?, timeoutMs?, render? (boolean|"auto"), waitMs? | | synoppy_screenshot | POST /api/screenshot | url, fullPage?, waitMs?, timeoutMs? | | synoppy_crawl | POST /api/crawl | url, limit? (1-25, default 10) | | synoppy_map | GET/POST /api/map | url | | synoppy_search | POST /api/search | query, maxResults?, markdown?, includeDomains?, excludeDomains?, fanout? | | synoppy_extract | POST /api/extract | url, prompt? (alias instruction?), schema? (JSON Schema → typed output) | | synoppy_classify | POST /api/classify | url, labels? (omit for NAICS/SIC) | | synoppy_enrich | GET/POST /api/brand | one of url, domain, or email | | synoppy_images | POST /api/images | url |

• synoppy_read — URL → clean markdown / HTML / text plus page metadata (title, description, language, statusCode, wordCount, rendered, bytesIn…). Pass render: "auto" to render JS-heavy pages on demand.
• synoppy_screenshot — full-page or viewport PNG as a data URL. Requires rendering credits; may return RENDER_UNAVAILABLE.
• synoppy_crawl — crawl a site, read each page (pages[], discovered, count).
• synoppy_map — every URL on a domain (source: "sitemap" | "links").
• synoppy_search — search the live web: ranked results with clean source URLs; markdown: true also reads each hit.
• synoppy_extract — AI structured-JSON extraction. Returns data, usage tokens, truncated, and model. Pass schema (a JSON Schema) for typed output constrained to that exact shape.
• synoppy_classify — industry (NAICS/SIC) by default, or custom labels mode ({ label, matched, confidence, reasoning }).
• synoppy_enrich — brand profile (name, logo, colors, fonts, socials, address) from a URL, bare domain, or work email.
• synoppy_images — every image with src, alt, width, height.

Not live yet (coming soon): /api/act. This is intentionally not exposed as a tool.

Setup

Claude Desktop — claude_desktop_config.json

{
  "mcpServers": {
    "synoppy": {
      "command": "npx",
      "args": ["-y", "@synoppy/mcp"],
      "env": { "SYNOPPY_API_KEY": "syn_xxxxxxxxxxxx" }
    }
  }
}

Cursor — ~/.cursor/mcp.json (same shape)

Create a key in the dashboard. The server reads SYNOPPY_API_KEY (auth header Authorization: Bearer syn_...) from its environment; set SYNOPPY_BASE_URL to point at a self-hosted instance.

Credits

Synoppy is metered. Every successful response carries creditsUsed (number) and creditsRemaining (number or null), and the tools surface them verbatim. For example, synoppy_read returns:

{
  "success": true,
  "metadata": { "title": "Example", "statusCode": 200, "wordCount": 412, "rendered": false, "bytesIn": 18233 },
  "markdown": "# Example...",
  "latencyMs": 284,
  "creditsUsed": 1,
  "creditsRemaining": 4999
}

Watch creditsRemaining to track your balance as you go.

Local dev

npm install
npm run build
SYNOPPY_API_KEY=syn_... npm start

The published package ships a single bundled, minified dist/index.js (with the #!/usr/bin/env node shebang preserved) plus generated dist/index.d.ts types — no readable source.

MIT licensed.