API & MCP

MCP server

Streamable HTTP at https://scrapeselect.com/mcp. Tools: search_providers, get_provider, compare_products, estimate_cost, recommend_scraping_stack and get_recent_changes. Example client configuration:

{
  "mcpServers": {
    "scrapeselect": {
      "type": "http",
      "url": "https://scrapeselect.com/mcp"
    }
  }
}

HTTP API

The full schema is at /openapi.json. The main endpoints:

• GET /api/v1/providers — all providers and products.
• GET /api/v1/providers/{slug} and GET /api/v1/products/{provider}/{product} — plans, limits, features and the sources behind them.
• POST /api/v1/estimate — monthly cost range per product and the cheapest plan that fits a workload.
• POST /api/v1/recommend — an architecture and the top three products, with Fit Scores, reasons, limitations and exclusions.
• POST /api/v1/compare — up to three products side by side.
• GET /api/v1/changes — confirmed pricing changes.

curl -s https://scrapeselect.com/api/v1/recommend \
  -H 'content-type: application/json' \
  -d '{
    "workload": {
      "successful_pages_per_month": 500000,
      "javascript_ratio": 0.3,
      "anti_bot_level": "medium",
      "target_country": "US",
      "peak_concurrency": 20
    },
    "budget_usd": 1000
  }'

Only successful_pages_per_month is required. Anything you leave out is filled in with a stated assumption, returned in the response so you can see what the number depends on. Errors use a stable code field; exceeding the rate limit returns HTTP 429 with Retry-After.

JSON and text twins

Every provider and product page has machine-readable twins: add .json or .txt to the URL, e.g. /providers/zyte.json. The change log is at /changes.json and /changes.txt. An overview for language models is at /llms.txt.

Using the data

Quote the source and check date of any number you pass on, and treat estimates as estimates — the provider's checkout is authoritative.