Google Open Sources Science Skills Toolkit for Agentic Research Workflows

GoogleGoogle

· Updated

Google open-sourced its Science Skills toolkit on GitHub to provide AI agents with grounded scientific data and improved token efficiency. The release includes specialized capabilities for genomics and structural biology, integrating over 30 databases including AlphaGenome and UniProt.

Google open-sourced its Science Skills toolkit on GitHub, a collection of agent capabilities designed to accelerate scientific discovery. The release provides structured instructions and scripts that extend AI agents with domain-specific grounding across genomics, structural biology, and cheminformatics.
License
Apache 2.0 and CC-BY 4.0
Installation
npx skills add
Primary databases
AlphaGenome, AlphaFold Database, UniProt, and more
Dependency manager
uv
Core capability
Genomics and structural biology agents

The toolkit enables agentic workflows with higher token efficiency (using fewer tokens per task) and better grounding (connecting outputs to verifiable sources). In real-world tests, Google used these skills to analyze AK2 mutations, performing complex structural analysis significantly faster than traditional methods by integrating insights from the Gemini for Science suite.

The bundle installs via npx skills add google-deepmind/science-skills/, or through the curated plugin in Google Antigravity. It uses the uv package manager for dependencies, and skills like AlphaGenome and OpenAlex need their own API keys to enable full functionality.

View the full update on github.com
Google AI Developers
Google AI Developers
@googleaidevs
X

Building autonomous agents for scientific discovery? 🧬🤖 @GoogleDeepMind Science Skills is now available on GitHub. We've open-sourced this specialized toolkit to accelerate your agentic workflows with scientific grounding and higher token efficiency. Download now ↓ https://t.co/cwp1HOeKvo

77retweets517likes
View on X

Still wondering? A few quick answers below.

Science Skills is an open-source collection of agent capabilities designed for scientific research tasks. Released by Google on GitHub, it provides structured instructions and scripts that extend AI agents with domain-specific grounding in genomics, structural biology, and cheminformatics to accelerate discovery workflows.

You can install the Science Skills bundle using the command npx skills add google-deepmind/science-skills/ in your terminal. For Google Antigravity users, the toolkit is available as a curated Science plugin that can be downloaded through the application's customization settings.

The toolkit uses the uv package manager to handle dependencies, which is installed automatically upon first use. While many skills are functional immediately, specific tools like AlphaGenome and OpenAlex require an API key to be configured in a local environment file to unlock full access.

Yes, Google has released the Science Skills toolkit under the Apache License 2.0 for software and the Creative Commons Attribution 4.0 International License for other materials. Some third-party data sources referenced within the skills may have their own specific licenses and terms of use.

Science Skills improve agentic workflows by providing better grounding in verifiable scientific data and achieving higher token efficiency. This allows agents to perform complex tasks, such as structural biology analysis, more reliably and at a lower computational cost than using general-purpose models alone.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Google

See all AI news & updates from Google →

Share this update