Model Context Protocol

The Scavio MCP server

Give Claude Code, Cursor and any MCP client live web data — Google search, social, e-commerce, travel, jobs, real estate, app stores, ad libraries and company filings — behind one API key. 191 tools across 31 platforms, and control over how many of them load into your context.

Get an API key Read the docs

50 free credits on signup. No card required.

How do I give Claude Code live web data?

Add an MCP server that exposes web-data tools, then let the agent call them. With Scavio that is one command — no local install, no OAuth, no per-platform login:

claude mcp add scavio --transport http \
  --url https://mcp.scavio.dev/mcp \
  --header "x-api-key: YOUR_SCAVIO_API_KEY"

Then ask the agent to call get_usage to confirm the connection. It is free and costs no credits.

Install

The hosted server needs nothing installed. Running it locally needs Node.js 20+ and no clone or build.

Claude Code (hosted)

Nothing to install. The hosted server is the fastest way in.

claude mcp add scavio --transport http \
  --url https://mcp.scavio.dev/mcp \
  --header "x-api-key: YOUR_SCAVIO_API_KEY"

Claude Code (local)

Runs on your machine over stdio. Requires Node.js 20+.

claude mcp add scavio \
  -e SCAVIO_API_KEY=YOUR_SCAVIO_API_KEY \
  -- npx -y @scavio/mcp-server

Cursor / Claude Desktop / Windsurf / VS Code

Any MCP-compatible client. Paste into the client's MCP config.

{
  "mcpServers": {
    "scavio": {
      "command": "npx",
      "args": ["-y", "@scavio/mcp-server"],
      "env": {
        "SCAVIO_API_KEY": "YOUR_SCAVIO_API_KEY"
      }
    }
  }
}

Tool definitions cost context. Choose what loads.

Every registered tool's schema is sent to the model in each session. Registering all 191 tools puts 262KB of tools/list — roughly 70k tokens — into the window before you have typed anything, and exceeds what some clients accept at all. So Scavio registers a curated set by default and lets you pick the rest with one environment variable.

SCAVIO_PLATFORMS	Registers	Tools	tools/list
(unset)	The default set plus Extract	106	102 KB
all	Every platform	191	262 KB
extract,sec,g2	Just those three platforms	11	20 KB
none	No platform tools, only get_usage	1	~1 KB

Sizes are measured, not estimated — npm run toolslist reproduces them.

# A Google-and-Extract agent carries ~20KB instead of 262KB
SCAVIO_PLATFORMS=google,extract npx -y @scavio/mcp-server

What the agent can fetch

Google SERP, AI Mode, Maps, Shopping, News, Trends, Flights, Hotels

Extract

Any URL to clean markdown, text or HTML

Social

YouTube, TikTok, Instagram, Reddit, X, LinkedIn, Threads, Kuaishou

E-commerce

Amazon, Walmart, eBay, Target, Home Depot, TikTok Shop

Travel & local

Booking, Airbnb, Tripadvisor, Yelp

Property

Zillow, Redfin

Jobs

Indeed, Glassdoor

Apps & reviews

App Store, Google Play, G2, Capterra

Ads & filings

Google Ads, Meta Ad Library, SEC, Companies House

Frequently asked questions

How many tools does the Scavio MCP server expose?+

191 across 31 platforms plus Extract. By default it registers 106 of them, because loading all 191 costs 262KB of tools/list - roughly 70k tokens of context in every session before you have typed anything, and more tools than some clients will accept at all. Set SCAVIO_PLATFORMS to load exactly the platforms you need.

Why does the tool count matter for an agent?+

Every registered tool's schema is sent to the model on each session, so a large server spends your context window before the conversation starts. 262KB of definitions is context you cannot use for the actual task. Scavio lets you pick the platforms you need, so a Google-and-Extract agent carries roughly 20KB instead of 262KB.

Do I need to install anything?+

No. The hosted server at https://mcp.scavio.dev/mcp needs no local install - pass your API key in the x-api-key header. If you would rather run it yourself, npx -y @scavio/mcp-server runs it over stdio with Node.js 20+ and no clone or build step.

Which clients does it work with?+

Any MCP-compatible client. Claude Code, Claude Desktop, Cursor, Windsurf and VS Code are the ones we document explicitly, and the JSON config shown above is the standard shape those clients share.

What can the agent actually fetch?+

Google search and the Google verticals, YouTube, TikTok, Instagram, Reddit, X, LinkedIn and Threads, Amazon, Walmart, eBay, Target and Home Depot, Booking, Airbnb, Tripadvisor and Yelp, Zillow and Redfin, Indeed and Glassdoor, the App Store and Google Play, G2 and Capterra, the Google and Meta ad libraries, SEC and Companies House filings, and Extract for turning any URL into clean markdown.

How is this billed?+

Per credit, on the same balance as the REST API - the MCP server is a different way into the same account, not a separate product. Most endpoints cost 1 credit. New accounts start with 50 free credits and no card is required.

Does it need OAuth or a browser login?+

No. One API key, passed as an environment variable locally or an x-api-key header against the hosted server. There is no OAuth dance and no per-platform login to maintain.

What happens when the credits run out?+

Calls return HTTP 402 with a billing URL in the body, so the failure is explicit rather than silent. Topping up needs no code or config change - the same key keeps working.

Can I limit which platforms load without editing code?+

Yes. SCAVIO_PLATFORMS takes a comma-separated allowlist of platform keys, matched case- and punctuation-insensitively, so meta-ads, metaads and META_ADS are the same thing and twitter resolves to x. An unrecognised key is logged to stderr and skipped rather than failing the server.

Is get_usage always available?+

Yes, on every configuration including SCAVIO_PLATFORMS=none. It is free, costs no credits, and is the call every install guide points at to confirm the server is wired up correctly.

Ready to connect your AI agents?

Get your API key in seconds. Search Google, Amazon, YouTube, Walmart, and more with 50 free signup credits.

Start Building Read the Docs