Stealth
- Anti-bot measures and CAPTCHAs handled per endpoint
- Fingerprinting, proxy rotation and retries tuned by us, not you
- A call that still fails is not billed
Blocking, browser rendering, proxies and parsing run on our side. You send one request with one key and get typed JSON back: 199 endpoints across 35 platforms, over REST or as MCP tools.
Last updated
Every feature serves one job: a platform's public data as typed JSON, from code or an agent, at a known price.
| Feature | What you get | Where it's documented |
|---|---|---|
| Endpoint catalogue | One endpoint per platform and data type, 199 in all | /scrapers/ |
| Typed JSON | The same field names on every call | Each endpoint's schema in the docs |
| MCP server | Every endpoint as an agent tool | MCP server |
| OpenAPI spec | A machine-readable API definition | GET /openapi.json |
| Free discovery calls | /me, /endpoints and /openapi.json cost 0 credits | Discovery endpoints |
| Cursor pagination | next_cursor passed back as cursor until it is null | Pagination |
| Per-call billing | Fixed credits per endpoint; failed calls free | /pricing/ |
| Back-pressure signals | 429 or 503 with Retry-After | Rate limits |
Each endpoint is one GET call with named inputs, such as /instagram-profile/?usernames=natgeo. No selectors, no URLs to build, no HTML. Browse all scrapers.
The same field names on every call. Code and agents read followersCount, score or price directly, because the schema is published before the call runs.
JSON over REST and MCP, in one envelope: {"success": true, "data": {…}}. To turn a single page into text or Markdown, use the free web tools.
// GET /instagram-profile/?usernames=mrbeast { "success": true, "data": { "profile": { "username": "mrbeast", "fullName": "MrBeast", "followersCount": 456000000, "postsCount": 350, "verified": true, "scraped_at": "2026-03-09T15:19:49Z" }, "next_cursor": null } }
Every endpoint ships as an MCP tool with a typed schema and a plain-English description, so the agent chooses the right scraper itself. One remote server at mcp.scrapier.io/mcp works with Claude Code, Cursor, Windsurf, Codex, VS Code and other MCP clients.
One command in Claude Code, Cursor or Windsurf.
Describe the data you need.
The agent calls the right tool. Billed per successful call, at the same price as REST.
// instagram_user_posts(username="natgeo", limit=20) { "username": "natgeo", "count": 20, "posts": [ { "id": "C9xQ2…", "likes": 412880, "comments": 2194 }, { "id": "C9vL8…", "likes": 298113, "comments": 1607 } ] }
Send your key in the X-API-Key header on every request. Keys never go in URLs.
GET /endpoints lists every endpoint with its inputs and cost_credit. Free.
Pass data.next_cursor back unchanged as cursor until it is null.
402 with a balance means out of credits. 429 or 503: wait Retry-After seconds. Retry 5xx with backoff, free; set timeouts of 210 s or more.
Import /openapi.json into Postman, Insomnia, Bruno or a code generator.
An n8n community node, plus Zapier and Make through their HTTP and webhook modules.
Scrapier runs browser rendering, proxy rotation and retry logic for each endpoint. You never configure a proxy or a headless browser: you receive JSON, or an error you are not billed for.
A fixed credit price per endpoint, per successful call, listed before you call it. Failed calls are free, an empty result page still costs, and MCP and REST cost the same. See pricing.
Six layers you would otherwise build and maintain yourself.
countryCode on Google Search and country_code on TikTok trending videos| Layer | Scrapier | DIY scraper |
|---|---|---|
| Anti-bot and CAPTCHAs | Handled per endpoint | Constant maintenance |
| Proxies | Included in the credit price | Separate vendor and bill |
| Parsing | Typed JSON per endpoint | Selectors that break on redesigns |
| Agent access | MCP server, one tool per endpoint, 199 in all | Write your own tool wrappers |
| Billing | Per successful call; failed calls free | Servers, proxies and engineering time |
What Scrapier does not do, so you can rule it in or out quickly.
Scrapier turns 199 endpoints across 35 platforms into typed JSON calls. Each has a published schema, runs over REST or MCP, and costs a fixed credit price per successful call. Failed calls cost nothing.
Yes. Every endpoint is an MCP tool on one remote server, priced the same as REST.
Typed JSON with a published schema per endpoint, wrapped in a success and data envelope.
No. Scrapier runs them per endpoint and bills only successful calls.
No. It covers its catalogue of 35 platforms. The free web tools convert single pages to text, Markdown or JSON.
One key, 199 endpoints, typed JSON back.