MCP Server
ScrapeBadger provides a hosted MCP (Model Context Protocol) server atmcp.scrapebadger.com that lets AI agents access all ScrapeBadger APIs through structured, typed tools.
What is MCP?
Model Context Protocol is an open standard by Anthropic for connecting AI models to external tools and data sources. Instead of writing code, your AI agent discovers and calls ScrapeBadger tools naturally during a conversation.Key Features
544 MCP tools
Every ScrapeBadger API endpoint is available as a typed MCP tool with structured parameters, annotations, and output schemas.
Zero Setup
Hosted at mcp.scrapebadger.com — connect Claude with one click, or add one config snippet to any other AI tool.
10 Platforms
Works with Claude, Cursor, Windsurf, VS Code, Gemini, ChatGPT, OpenClaw, Cline, and more.
Same Pricing
No extra cost — MCP uses the same credit-based pricing as the REST API.
50 Products
Twitter/X, Instagram, Facebook, LinkedIn, Google, YouTube, TikTok, Reddit, Amazon, eBay, Walmart, Vinted, Depop, Wallapop, Gumtree, real estate, search engines, LLM answers, and general web scraping — all in one server.
Latest MCP Standard
Speaks the stateless 2026-07-28 protocol revision and every earlier one; OAuth 2.1 with PKCE, resource indicators, and Client ID Metadata Documents.
Available Tools
Google Tools
Google MCP tools are prefixed withgoogle_ — highlights:
- Search & Discovery
- Maps
- Shopping
- News, Trends, Jobs
- Specialized
Quick Start
Using Claude? No API key needed:1
Add the Connector
In claude.ai (or Claude Desktop), go to Settings > Connectors > Add custom connector and enter
https://mcp.scrapebadger.com/mcp.2
Sign In and Approve
Click Connect — you’ll be redirected to sign in with your ScrapeBadger account and approve access. That’s the whole setup.
3
Start Using Tools
Ask Claude to search tweets, scrape websites, or look up products — it discovers and calls the right tools automatically.
Authentication
Two authentication modes are supported:OAuth (recommended)
Clients that support MCP OAuth can connect without any key handling: add the server URL, sign in with your ScrapeBadger account, and approve the consent screen. This includes claude.ai, Claude Desktop, Claude Code, ChatGPT, Codex, Cursor, VS Code + Copilot, Gemini CLI, Zed, Microsoft 365 Copilot and Perplexity. The consent screen shows which app is asking and where you’ll be sent back to. Access is tied to your account’s API key and can be revoked anytime by deactivating that key in the dashboard.API key
Protocol & Standards
The server tracks the latest MCP specification (revision 2026-07-28):- Protocol versions — speaks the stateless 2026-07-28 revision (per-request versioning,
server/discover) and serves every earlier revision (2025-11-25, 2025-06-18, …) from the same endpoint, so older clients keep working. Streamable HTTP transport only; the deprecated HTTP+SSE transport is not offered. - Tool metadata — every tool carries a human-readable
title,readOnlyHint/destructiveHint/openWorldHintannotations, and an output schema. Tool results are size-capped to stay within client limits, and the tool list is served with cache hints (ttlMs/cacheScope) for faster clients. - OAuth 2.1 — authorization code + PKCE (S256), refresh-token rotation, RFC 8414 authorization-server metadata, RFC 9728 protected-resource metadata, RFC 8707 resource indicators, RFC 9207
issauthorization responses, and Client ID Metadata Documents (HTTPS-URL client ids, the 2026-07-28 successor to Dynamic Client Registration — which remains supported for older clients). Loopback redirect URIs match on any port (RFC 8252 §7.3), so native clients that pick a port at sign-in time work, and DCR clients that ask forclient_secret_basic/client_secret_postreceive a client secret.
MCP Endpoint
https://mcp.scrapebadger.com/health
Tool discovery and toolsets
Listing every tool up front would put ~115k tokens of tool schemas into your agent’s context before it reads your message. Some clients also cap the number of tools they pass on (Cursor reportedly at 40). So by default the endpoint lists only eight small helper tools. Your agent reaches every endpoint through them:search_endpoints("google maps reviews")finds the tool.describe_endpoint("google_maps_reviews")returns its parameters, credit cost and latency.call_endpoint("google_maps_reviews", {...})runs it. Usejq,fieldsormax_itemsto trim the response.
quote_call, get_request, get_account_info, report_bug and send_feedback. All are free.
To have specific APIs’ tools listed directly, add apis to the URL:
Use the API names
search_endpoints returns (google, google-play, amazon, twitter, …). Every tool can be called by name on any URL, even when it isn’t listed. Older clients and saved prompts that call amazon_search directly keep working.
Lean endpoint for agents
https://mcp.scrapebadger.com/mcp/lean serves only the helper tools, for clients that load every tool definition into context:
search_endpointsanddescribe_endpointquote_callandcall_endpointget_requestreport_bugandsend_feedbackget_account_info
call_endpoint. Both endpoints return oversized results as a preview; call again with jq, fields or max_items to get the part you need (the new call is billed, because results are not stored). Long calls return a request_id. See AI Agents.
Vinted mobile reads
Callvinted_mobile_operations to discover parameters and examples, then call vinted_mobile_read with operation, market, and a parameters object. See the mobile API catalog for pricing, sold comparables, seller reviews, shipping, and reference data.
