ADK Agent Directory

What is this project?

A curated directory of AI agents powered by Google's Agent Development Kit (ADK) and the Gemini 2.5 Flash model. The directory serves as a centralized hub where developers can discover, share, and contribute specialized AI agents that help with tasks like web search, image generation, document processing, and more. It's an open-source, community-driven platform designed to make AI agents more accessible and discoverable.

Why did I build it?

I wanted to create a centralized resource for the ADK community where developers could easily discover and share their AI agents. Building agents is one thing, but making them discoverable and usable by others is a different challenge. The directory addresses this by providing a well-documented, searchable platform where each agent includes metadata, sample prompts, use cases, and documentation—making it easy for others to understand and integrate agents into their own projects.

How does it work?

The directory operates as a community-driven platform with a structured contribution workflow:

1. Agent Discovery: Users can browse agents by category, tags, or search functionality
2. Agent Details: Each agent includes comprehensive metadata:
   • Name, display name, and description
   • Supported tools and capabilities
   • Tags and use cases
   • Sample prompts and examples
   • Author information and documentation links
3. Community Contributions: Developers can contribute by:
   • Forking the repository
   • Building agents following ADK best practices
   • Creating a metadata.json file with agent details
   • Submitting a pull request for review

Agents in the directory leverage ADK's capabilities including tool-calling abilities to interact with external services, session context for maintaining conversation memory, and artifact handling for outputs like images and documents.

Tech Stack

• Backend: Google ADK, Python, FastAPI, PostgreSQL/Neon
• Frontend: Next.js 16, React 19, Tailwind CSS v4, Material Design 3
• LLM: Gemini 2.5 Flash
• Database: PostgreSQL/Neon for agent metadata and user data

What were the biggest challenges?

• Standardizing Agent Metadata: Creating a flexible yet consistent metadata schema that captures all necessary information while remaining easy for contributors to fill out
• Community Engagement: Building a contribution workflow that's accessible to developers while maintaining quality standards
• Discoverability: Designing search and filtering systems that help users find the right agents for their specific needs
• Documentation Quality: Ensuring each agent has sufficient documentation, examples, and use cases to be truly useful

How can you use it or build on this?

• Browse the directory to discover agents for specific tasks
• Use agents directly in your projects by following their documentation
• Contribute your own agents to help the community
• Learn from existing agents to understand ADK best practices
• Extend the platform with additional features like agent ratings, usage analytics, or integration guides

Links

• Website: https://agentdirectory.folch.ai
• About: https://agentdirectory.folch.ai/about
• Contribute: https://agentdirectory.folch.ai/contribute

What's next?

• Enhanced search and filtering capabilities
• Agent usage analytics and popularity metrics
• Integration guides and code examples for each agent
• Agent versioning and update notifications
• Community ratings and reviews system
• Automated testing and validation for contributed agents

In the age of AI, stay curious!