autogen-scavio

AutoGen integration for the Scavio API, a search API for AI agents. Provides 188 tools across 31 platforms -- retail, real estate, travel, jobs, app stores, company filings, ad libraries and social -- plus scavio_extract, which reads any URL as clean Markdown.

Installation

pip install autogen-scavio

Setup

Get a free API key at dashboard.scavio.dev and set it as an environment variable:

export SCAVIO_API_KEY="sk_live_..."

Quick Start

import asyncio
from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_scavio import create_scavio_tools

tools = create_scavio_tools()

agent = AssistantAgent(
    name="researcher",
    model_client=OpenAIChatCompletionClient(model="gpt-5.5"),
    tools=tools,
    system_message="You are a research assistant with access to search tools.",
)

async def main():
    result = await agent.run(task="What are the top AI agent frameworks in 2026?")
    print(result.messages[-1].content)

asyncio.run(main())

Selective Tool Loading

Load only the providers you need:

# Only Google and Amazon tools
tools = create_scavio_tools(providers=["google", "amazon"])

# Only YouTube
tools = create_scavio_tools(providers=["youtube"])

# Retail price comparison across four marketplaces
tools = create_scavio_tools(providers=["amazon", "walmart", "ebay", "target"])

# Read a page no platform endpoint covers
tools = create_scavio_tools(providers=["extract"])

# Providers: extract, google, amazon, youtube, walmart, reddit, tiktok,
# tiktok_shop, instagram, x, linkedin, threads, kuaishou, ebay, target,
# home_depot, zillow, booking, tripadvisor, indeed, airbnb, glassdoor, yelp,
# app_store, google_play, sec, redfin, companies_house, g2, capterra,
# google_ads, meta_ads

# Individual tool factories
from autogen_scavio import create_web_search_tool, create_extract_tool

web_tool = create_web_search_tool(max_results=10)
extract_tool = create_extract_tool()

Amazon changed in 0.4.0 (breaking). The upstream provider moved. domain is replaced by country, a two-letter marketplace code (us, gb -- the UK is gb, not uk -- de, jp, ...). sort_by, pages, category_id, merchant_id, language, currency, device, zip_code and autoselect_variant are gone: the marketplace ignores all of them, and sort_by was verified to return the identical unordered set for every value, so they are removed rather than kept as silent no-ops. Responses are normalized -- price is a number with a sibling currency, and buybox is gone (use scavio_amazon_offers).

Google v1 is gone (0.5.0). /api/v1/google was retired on 2026-08-04 and now answers 410, so every Google tool calls a /api/v2/google endpoint and takes v2 params natively: gl, hl, start, google_domain, device. country_code, language and page no longer exist and are not mapped -- start is a zero-based result offset (0, 10, 20), not a 1-based page number, so a silent remap would fetch the wrong page. Google responses are flat: the payload sits at the top level, there is no data wrapper, and organic hits carry link / snippet, not v1's url / content.

Reddit is complete in 0.6.0. All twelve endpoints are now tools; only search and post detail existed before. Three shapes to keep straight when chaining them: scavio_reddit_search returns data.results, scavio_reddit_post returns a flat post object with no comments at all, and the subreddit, user and popular feeds return data.posts. Read a thread with scavio_reddit_post_comments, then expand a subtree by passing that comment's reply_cursor to scavio_reddit_comment_replies. Sort values are uppercase (HOT, NEW, TOP, BEST, CONTROVERSIAL); the subreddit feed is the only one that also takes RISING.

0.7.0 adds 91 tools: 21 new platforms, 5 more Walmart endpoints, and scavio_extract. Requires scavio>=0.15.0 -- an older SDK has none of the new namespaces and every tool below fails at import. Three things worth knowing before you wire them up:

• scavio_extract is the fallback for everything else. It reads any URL and returns clean Markdown (or plain text, or raw HTML), so an agent that hits a page no platform endpoint covers still has a move. It leads the default tool set for that reason.
• Five platforms are lookup-first. Tripadvisor, Glassdoor, SEC EDGAR, Companies House and Google Ads Transparency are keyed by ids that only exist inside their own URLs, so a caller holding a name must resolve it first: scavio_tripadvisor_locations, scavio_glassdoor_companies, scavio_sec_lookup, scavio_companies_house_search, scavio_google_ads_advertisers.
• Four surfaces are priced by the request body, not the route. Walmart by domain, Threads by whether you pass user_id or username, Kuaishou per endpoint (1 to 40), and extract by mode. Every tool description states its own price; see Credits.

These tools return the API response verbatim -- max_results does not apply to them. Each endpoint pages itself (page, cursor, per_page, limit, count), and trimming rows out of a response that still reports has_more would read as an empty tail.

Available Tools

Credits

Every tool states its own cost in its description, so the agent sees the price before it picks. Most endpoints are 1 credit; these are the exceptions:

All 14 Google tools, all 11 X tools, the remaining LinkedIn and Kuaishou tools and everything under Amazon, Reddit, TikTok, TikTok Shop, eBay, Target, Zillow, Redfin, Booking.com, Airbnb, App Store, SEC EDGAR, Companies House, Google Ads Transparency and Meta Ad Library is 1 credit.

Four surfaces are priced by the request body, so no single number is true for them:

Five LinkedIn endpoints -- person/contact, company/people, company/jobs, search/people and search/posts -- were retired upstream and only answer 410, so they are deliberately not wrapped as tools: a tool that always fails is still a menu item a model will pick. Use scavio_linkedin_company's featured_employees in place of company/people, and scavio_linkedin_search_jobs with a company name in place of company/jobs.

create_youtube_metadata_tool is still exported but is a deprecated alias of scavio_youtube_video (same endpoint, same response), so it is no longer part of the default tool set. Use the video tool.

TikTok Shop caveats

tools = create_scavio_tools(providers=["tiktok_shop"])

Two limits are worth knowing before chaining these tools:

• scavio_tiktok_shop_product resolves only about 44% of the product ids that scavio_tiktok_shop_search returns. Upstream has no detail data for the rest, so a not-found result is a normal outcome rather than an error -- skip the product instead of retrying. Search is a listing source, not the first leg of a reliable search-then-detail pipeline.
• scavio_tiktok_shop_product does not return a price. Upstream masks the digits on the product page, so price.current and price.original come back null. Exact prices are on scavio_tiktok_shop_search, scavio_tiktok_shop_shop_products and scavio_tiktok_shop_category_products.

Configuration

from autogen_scavio import configure

# Set API key programmatically
configure(api_key="sk_live_...")

# Or pass to factory
tools = create_scavio_tools(api_key="sk_live_...", max_results=10)

About Scavio

Scavio is a unified search API built for AI agents — one API key, structured JSON, no scraping or proxies. A real-time Tavily alternative and SerpAPI alternative with data from:

• Google Search API — SERP results, news, images, maps, and knowledge graph
• Amazon Product API and Walmart Product API — product search and details
• eBay API, Target API and Home Depot API — retail listings, product detail and reviews
• Zillow API and Redfin API — listings, property detail and market stats
• Booking.com API, Airbnb API and Tripadvisor API — stays, rates and guest reviews
• Indeed API and Glassdoor API — job postings, employer profiles, reviews and salaries
• G2 API and Capterra API — B2B software ratings and review bodies
• App Store API and Google Play API — app listings, install counts and reviews
• SEC EDGAR API and Companies House API — US filings, XBRL facts and the UK register
• Google Ads Transparency API and Meta Ad Library API — live competitor creatives
• YouTube API, TikTok API, and Instagram API — video and social media data
• TikTok Shop — product detail, reviews, categories, and seller catalogs
• Threads API and Kuaishou API — profiles, posts and search
• Reddit API — posts and threaded comments
• X (formerly Twitter) and LinkedIn — profiles, posts, and search
• Extract — read any URL as clean Markdown, plain text or raw HTML

Get a free API key and explore the documentation. You can also compare Scavio vs alternatives on coverage and pricing.