anakin-adk

Google ADK tools for web scraping, search, and research — powered by Anakin.

Give your Gemini agents the ability to scrape websites, search the web, and run deep research — all through a simple toolkit that plugs into Google's Agent Development Kit.

Installation

pip install anakin-adk

You also need the Anakin CLI installed and authenticated:

pip install anakin-cli
anakin login --api-key "ask_your-key-here"

Get your API key by signing up at anakin.io/signup — then grab the key (starts with ask_) from your dashboard. Verify with:

anakin status

Quick start

from anakin_adk import AnakinToolkit
from google.adk.agents import Agent

agent = Agent(
    model="gemini-2.5-pro",
    name="web_researcher",
    instruction="Help users extract data from the web",
    tools=AnakinToolkit().get_tools(),
)

Run with the ADK dev UI:

adk web

How it works

Your ADK Agent (Gemini)
        │
        ▼
   anakin-adk          ← this package: exposes tools to ADK
        │
        ▼
   anakin-cli           ← local CLI that talks to the Anakin API
        │
        ▼
   Anakin Cloud API     ← handles scraping, search, and research

1. Your Gemini agent decides to call a tool (e.g. scrape_website).
2. anakin-adk translates the call into an anakin-cli command.
3. The CLI sends the request to the Anakin cloud API, which handles proxy rotation, browser rendering, and anti-detection.
4. Results come back as clean markdown or structured JSON, ready for the agent to use.

Available tools

Using individual tools

You can pick only the tools you need:

from anakin_adk import ScrapeWebsiteTool, SearchWebTool
from google.adk.agents import Agent

agent = Agent(
    model="gemini-2.5-pro",
    name="search_and_scrape",
    instruction="Search the web and scrape relevant pages",
    tools=[SearchWebTool(), ScrapeWebsiteTool()],
)

Examples

See the examples/ directory:

• basic_scraping.py — Simple scrape agent
• research_agent.py — Deep research agent
• search_and_scrape.py — Multi-step: search then scrape

Troubleshooting

For JS-heavy sites that return empty content, try setting use_browser=true in the scrape_website tool.

Check your auth status anytime with anakin status.

Development

pip install -e ".[dev]"
pytest tests/ -v
ruff check anakin_adk/

License

MIT