google-adk-scavio

Scavio real-time search and page-extraction tools for Google ADK (Agent Development Kit) — Google, YouTube, Amazon, Walmart, Reddit, scavio_extract (read any URL), and opt-in website tools that map, crawl and batch-extract whole sites, with one API key.

New in 0.4.0: map, crawl and batch (opt-in). create_scavio_tools(enable_crawl=True) adds scavio_map_website, scavio_crawl_website, scavio_crawl_status, scavio_batch_extract and scavio_batch_status. Requires scavio>=0.23.0. See Map, crawl and batch.

Amazon changed (breaking). The upstream provider moved in 2026-07: domain is replaced by country, a two-letter marketplace code (us, gb -- the UK is gb, not uk -- de, jp, ...), and sort_by, pages, category_id, merchant_id, language, currency, device, zip_code and autoselect_variant are gone. The marketplace ignores all of them (sort_by returns the identical unordered set for every value), so they are removed rather than kept as silent no-ops. Rank and filter results yourself.

Reddit changed (breaking). Reddit moved upstream in 2026-07 and now costs 1 credit, not 2. scavio_reddit_search takes query and cursor only: type and sort never reached the provider and are gone rather than kept as silent no-ops. scavio_reddit_post returns a flat post object and no comments -- fetch those from the Reddit comments endpoints via MCP or the scavio SDK.

Install

pip install google-adk-scavio

Setup

Get a Scavio API key from the Scavio Dashboard (new accounts get 50 free signup credits, no credit card). Set SCAVIO_API_KEY or pass api_key= to the factory.

Usage

from google.adk.agents import Agent
from google_adk_scavio import create_scavio_tools

root_agent = Agent(
    name="search_assistant",
    model="gemini-2.5-flash",
    instruction="Search the web, news, shopping sites, and Reddit with the Scavio tools.",
    tools=create_scavio_tools(),  # reads SCAVIO_API_KEY
)

Expose only the tools you need:

tools = create_scavio_tools(
    enable_extract=True,
    enable_google=True,
    enable_reddit=True,
    enable_youtube=False,
    enable_amazon=False,
    enable_walmart=False,
)

enable_extract is the odd flag out: every other flag gates a platform, while scavio_extract is a core endpoint that reads any URL and belongs to none. enable_crawl (off by default, because a crawl bills per page) adds the five website tools.

Tools

create_scavio_tools() returns plain typed Python functions; ADK wraps each one in a FunctionTool automatically. Each returns the structured Scavio JSON response as a dict; errors come back as {"error": "..."} rather than raising, so a failed call never crashes the run.

Reading a page

scavio_extract is the one tool here that is not tied to a platform: give it any URL and it returns the page as data.content, alongside data.url, data.format, data.mode and data.content_length.

Its price is a function of mode, not a per-call constant. normal (the default) is the standard fetch and costs 1 credit, advanced renders JavaScript-built pages and costs 1, and ultra reaches the hardest pages and costs 2. Start on normal. Only a successful read is billed — a dead link, a bot wall or a timeout costs nothing. See the extract docs.

format picks the shape: markdown (the default) is a readability extraction with the page chrome stripped, text is that flattened to plain text, and html is the raw page.

Map, crawl and batch

When one page is not enough, enable_crawl=True adds five website tools. scavio_map_website answers in one call. scavio_crawl_website and scavio_batch_extract start a background job, wait up to wait_seconds (default 45, max 300) for it to finish, and return the job's status with its first page of results (10 by default) either way. A job that is still queued or running when the wait ends keeps going: the agent reads the rest, free, with scavio_crawl_status / scavio_batch_status and the job id. That keeps an agent turn bounded while a large crawl (up to 6 hours) runs on.

root_agent = Agent(
    name="docs_reader",
    model="gemini-2.5-flash",
    instruction="Map the site first, then crawl only the section you need.",
    tools=create_scavio_tools(enable_crawl=True, enable_youtube=False, enable_amazon=False,
                              enable_walmart=False, enable_reddit=False),
)

# Or call the functions directly:
tools = {t.__name__: t for t in create_scavio_tools(enable_crawl=True)}
tools["scavio_map_website"]("https://hono.dev/docs", limit=50)
job = tools["scavio_crawl_website"]("https://hono.dev/docs", max_pages=25, include_paths=["/docs/**"])
tools["scavio_crawl_status"](job["data"]["id"], cursor=job["data"]["next_cursor"])
tools["scavio_batch_extract"](["https://example.com/", "https://hono.dev/docs"])

Pricing. Map costs 1 credit in normal mode (advanced up to 2, ultra up to 10, priced by what the call needed). Crawl and batch bill per successful page at extract's price for the mode -- normal and advanced 1 credit, ultra 2 -- while failed and skipped pages are free; if your balance runs out the job stops cleanly with stopped_reason: "insufficient_credits". Reading a job is free. Crawl and map obey the site's robots.txt; results are kept for 3 days. Plan limits (pages per crawl / active jobs / URLs per batch): Free 10/1/10, PAYG 100/1/50, Project 500/2/100, Bootstrap 2,000/3/500, Startup 5,000/5/1,000, Growth 10,000/10/1,000. Webhooks, idempotency keys, cancellation and the rarer crawl options (subdomains, sitemap mode, query-string handling) are on the REST API and the scavio SDK (client.extract.crawl(...), client.extract.cancel_crawl(...)), not exposed as agent tools.

Use every endpoint via MCP

This package is a curated subset of the Scavio API, not the whole thing: 21 tools — 15 across 5 platforms, plus the core extract endpoint and the 5 opt-in website tools. For full coverage — all 195 endpoints across 31 platforms (Google, YouTube, Amazon, Walmart, Reddit, TikTok, TikTok Shop, Instagram, X, LinkedIn, eBay, Target, Home Depot, Zillow, Redfin, Booking, Airbnb, Tripadvisor, Yelp, Indeed, Glassdoor, the App Store, Google Play, SEC, Companies House, G2, Capterra, Google Ads, Meta Ad Library, Threads and Kuaishou) — point ADK at the hosted MCP server at mcp.scavio.dev, which exposes 191 tools. Requires ADK's MCP extra: pip install "google-adk[mcp]".

from google.adk.agents import Agent
from google.adk.tools.mcp_tool import McpToolset, StreamableHTTPConnectionParams

scavio = McpToolset(
    connection_params=StreamableHTTPConnectionParams(
        url="https://mcp.scavio.dev/mcp",
        headers={"x-api-key": "sk_live_..."},
    )
)

root_agent = Agent(name="search_assistant", model="gemini-2.5-flash", tools=[scavio])

Links

• Scavio: https://scavio.dev
• Docs: https://scavio.dev/docs/google-adk
• Dashboard: https://dashboard.scavio.dev

License

MIT