Arcade.dev Integration
Integrate Scavio with Arcade.dev to give your agents real-time web search across Google, Google News, Reddit, YouTube, and Amazon -- delivered as first-class Arcade/MCP tools with securely managed secrets.
Secrets stay secret
SCAVIO_API_KEY at call time and hosts it encrypted -- the client and the LLM never see the value.Introduction
The arcade-scavio toolkit is built on arcade-mcp-server. Run it standalone over stdio/HTTP, or deploy it to Arcade Cloud. It is a cost-effective Tavily and SerpAPI alternative.
Step-by-Step Integration Guide
Step 1: Install the toolkit
pip install arcade-scavioStep 2: Set your API key
Get a key at dashboard.scavio.dev, then expose it as the SCAVIO_API_KEY secret:
export SCAVIO_API_KEY=sk_live_...Step 3: Run the server
python -m arcade_scavio.server # stdio (Claude Desktop, CLI tools)
python -m arcade_scavio.server http # HTTP+SSE (Cursor, VS Code)Available Tools
| Tool | Description |
|---|---|
search_google | Google SERP -- real-time organic web results |
search_news | Google News -- recent articles on a topic |
search_reddit | Reddit posts -- community discussion and sentiment |
search_youtube | YouTube -- videos |
search_amazon | Amazon -- product listings |
map_website | A site's page URLs from its sitemaps and start page |
crawl_website | Crawl a site from one URL -- every page as Markdown, text or HTML |
get_crawl_status | A crawl's progress and page results, cursor-paged (free) |
batch_extract | Extract a list of URLs in one job |
get_batch_status | A batch's progress and page results, cursor-paged (free) |
The five website tools ship in arcade-scavio 0.3 and later. Crawl and batch start a background job, wait up to wait_seconds (default 45, max 300) and return the job id with the first page of results; a job still running when the wait ends keeps going, and the status tools read the rest by id and cursor. Map is 1 credit in normal mode. Crawl and batch bill per successful page at Extract's price for the mode (1, 1 or 2 credits); failed and skipped pages are free, and the status tools cost nothing. Crawl and map obey the site's robots.txt; results are kept for 3 days. API reference: map, crawl, crawl status, batch and batch status; plan limits are on the crawl and batch pages, and single-page reads use Extract.
Deploy to Arcade Cloud
Package and host the toolkit so the Arcade Engine can fulfil tool calls for your users:
arcade login
arcade deployArcade discovers the SCAVIO_API_KEY secret and stores it encrypted.
Prefer the hosted MCP server?
Scavio also runs a hosted MCP server at mcp.scavio.dev with the full tool catalog (Walmart, TikTok, Instagram, Maps, Shopping, and more) -- no install.
Benefits of Scavio + Arcade.dev
- Secure secrets: keys injected at call time, never exposed.
- MCP-native: run standalone or on Arcade Cloud.
- Multi-platform: web, news, social, shopping in one toolkit.
- Cost-effective: most calls cost a single credit.
Next Steps
- MCP Integration -- the full tool catalog
- Google Search API -- endpoint reference