MCP Server
Connect your AI assistant to 20+ scraping and data tools with one command.
Overview
The Model Context Protocol (MCP) is a standard that lets AI assistants call external tools directly. Scraper exposes all its tools as an MCP server, so your assistant can search Google, fetch reviews, find jobs, and more without you writing any integration code.
No code needed
Connect once, then ask your assistant in natural language. It picks the right tool automatically.
Secure by default
OAuth 2.1 with PKCE handles authentication. No API keys to copy or manage.
Same billing
Uses the same credit system as the REST API. No extra fees for MCP access.
Setup
On first use, your client opens a browser window to authorize. After that, the token refreshes automatically.
Claude Code
claude mcp add scraper --transport http https://scraper.run/api/mcp
Claude Desktop
Add to your claude_desktop_config.json:
{"mcpServers": {"scraper": {"type": "streamable-http","url": "https://scraper.run/api/mcp"}}}
Cursor
Add to .cursor/mcp.json in your project:
{"mcpServers": {"scraper": {"type": "streamable-http","url": "https://scraper.run/api/mcp"}}}
VS Code
Add to your .vscode/mcp.json:
{"servers": {"scraper": {"type": "http","url": "https://scraper.run/api/mcp"}}}
Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
{"mcpServers": {"scraper": {"type": "streamable-http","url": "https://scraper.run/api/mcp"}}}
Other clients
Any MCP client that supports Streamable HTTP can connect. Point it at https://scraper.run/api/mcp. OAuth metadata is at https://scraper.run/api/.well-known/oauth-authorization-server
Authentication
The MCP server supports two authentication methods:
| Method | How it works | Best for |
|---|---|---|
| OAuth 2.1 | Browser-based consent flow with PKCE, automatic token refresh | Claude, Cursor, VS Code, Windsurf |
| API Key | Pass your sc_live_ key as a Bearer token | Programmatic or headless clients |
OAuth tokens have a 1-hour lifetime and refresh automatically. API keys work the same as with the REST API.
Available Tools
All tools use the same credit system as the REST API.
| Tool | Description | Credits |
|---|---|---|
| google_search | Web search via Google | 0.5 |
| google_translate | Translate text between languages | 1 |
| google_vision | Analyze images (labels, text, landmarks) | 0.5 |
| google_geocode | Geocode addresses or reverse-geocode coordinates | 0.1 |
| google_places_search | Search for places by query | 1 |
| google_places_details | Get place details by place ID | 0.6 |
| google_places_nearby | Find places near a location | 1 |
| google_places_autocomplete | Autocomplete place names | 0.1 |
| google_maps_reviews | Get Google Maps reviews for a place | 1 |
| trustpilot_reviews | Get Trustpilot reviews for a business | 1 |
| tripadvisor_reviews | Get Tripadvisor reviews | 1 |
| yelp_reviews | Get Yelp reviews for a business | 1 |
| glassdoor_reviews | Get Glassdoor company reviews | 1 |
| glassdoor_company | Get Glassdoor company overview | 1 |
| glassdoor_salaries | Get salary data from Glassdoor | 1 |
| glassdoor_interviews | Get interview experiences | 1 |
| glassdoor_jobs | Search Glassdoor job listings | 1 |
| glassdoor_benefits | Get company benefits info | 1 |
| indeed_jobs | Search Indeed job listings | 1 |
| tiktok_product | Get TikTok Shop product details | 1 |
| tiktok_product_reviews | Get TikTok Shop product reviews | 1 |
| youtube_search | Search YouTube videos, channels, playlists | 1 |
| youtube_video | Get metadata for a YouTube video | 0.5 |
| youtube_channel | Get metadata for a YouTube channel | 0.5 |
| youtube_channel_videos | List videos for a YouTube channel | 1 |
| youtube_playlist | Get metadata and videos for a YouTube playlist | 0.5 |
| youtube_hashtag | Find YouTube content by hashtag | 0.5 |
| youtube_transcript | Get the transcript for a YouTube video | 0.5 |
| youtube_comments | Get comments on a YouTube video | 1 |
| reddit_post | Get a Reddit post and its comments | 1 |
| reddit_search | Search Reddit | 1 |
| hackernews_post | Get a HackerNews post and its comments | 1 |
| hackernews_search | Search HackerNews | 1 |
| booking_hotel | Get full hotel details from Booking.com | 1 |
| booking_reviews | Get guest reviews for a Booking.com hotel | 1 |
| booking_search | Search Booking.com hotels by destination | 1 |
| g2_analytics | Get aggregated analytics for G2 products | 1 |
| g2_products | Search or look up G2 software products | 1 |
| g2_discussions | Fetch Q&A discussion threads for a G2 product | 1 |
| g2_reviews | Get G2 software reviews | 1 |
| twitter_thread | Get a Twitter/X post and its replies | 2 |
| twitter_search | Search Twitter/X for recent tweets | 2 |
| twitter_user_tweets | Get recent tweets from an X/Twitter user | 2 |
| twitter_user_info | Get an X/Twitter user profile | 1 |
| linkedin_post | Get a LinkedIn post and its comments | 2 |
| instagram_post | Get an Instagram post and its comments | 2 |
| instagram_search | Search Instagram posts by hashtag | 2 |
| threads_post | Get a Threads post and its replies | 2 |
| threads_search | Search Threads for recent posts | 2 |
| facebook_post | Get a Facebook post and its comments | 5 |
| facebook_search | Search Facebook for posts and pages | 5 |
Selecting fields
Most tools accept an optional fields parameter that limits the response to specific fields. Use it to reduce token cost when you only need part of the response. Click any tool name above to see its full field reference.
Example: passing fields: ["rating", "text", "author"] to trustpilot_reviews returns only those three fields per review.
Examples
Once connected, ask your assistant in natural language. It picks the right tool automatically.
Single tool
“Get Trustpilot reviews for notion.so and list the most common complaints”
“Find senior React developer jobs on Indeed in San Francisco”
“What are the Google Maps reviews for Blue Bottle Coffee in Tokyo?”
“Translate this paragraph to Japanese”
Multi-tool chains
“Search Google for Notion competitors and get Trustpilot reviews for each one”
“Find React developer jobs on Indeed in San Francisco, then translate the top 3 descriptions to Vietnamese”
“Compare Glassdoor reviews and salaries for Meta vs Google for senior engineers”
“Find coffee shops near Times Square and get their Google Maps reviews”
Troubleshooting
My assistant uses its built-in web search instead of google_search
Some assistants (like Claude) have a native web search tool. For generic search prompts, they may prefer their own tool over the MCP version. Two workarounds:
- Chain it with another Scraper tool so the assistant uses both in one flow (e.g. “Search Google for X and get Trustpilot reviews for each result”)
- Reference the tool by name: “Use the google_search tool to find...”
OAuth flow not opening in the browser
Make sure your client supports Streamable HTTP transport. Some older MCP clients only support stdio. Check that you're using the latest version of your client.
Session expired or 401 errors
OAuth tokens expire after 1 hour and should refresh automatically. If you see repeated 401 errors, remove and re-add the MCP server to trigger a fresh authorization flow.