RabbitHole is a Python framework for building and orchestrating LLM-powered agents. It acts as a universal translator, providing a standardized Agent-to-Agent (A2A) communication protocol that allows agents from different providers (Google, OpenAI, etc.) to work together seamlessly.
While inspired by tools like LangChain that help build powerful individual agents, RabbitHole focuses on a higher level of abstraction: multi-agent orchestration. It provides the backbone for a society of agents to collaborate on complex tasks.
Building AI agents involves more than just writing prompts. You need to handle server setup, communication protocols, and task management. RabbitHole abstracts away this boilerplate, providing a robust foundation for your agent applications.
Use RabbitHole for:
- Standardized Communication: Built on Google's A2A protocol, RabbitHole ensures your agents speak a common language, making them interoperable by design.
- Rapid Development: Get an agent server running with minimal code. Our framework provides the structure, so you can focus on your agent's unique logic.
- Extensible by Design: With a plug-in architecture, RabbitHole makes it easy to integrate agents from different providers. We currently support Google's ADK and OpenAI's Agent SDK, with more on the way.
- Integrated Tooling: Comes with a built-in CLI and a Streamlit UI for easy testing and interaction with your agents right out of the box.
Get started by installing RabbitHole directly from the source. We recommend using a virtual environment.
# 1. Clone the repo
git clone https://github.com/VinsmokeSomya/RabbitHole.git
cd RabbitHole
# 2. Create a virtual environment and activate it
# We recommend `uv`, but Python's venv works great too.
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
# 3. Install in editable mode with all dev dependencies
pip install -e ".[dev]"Let's run the pre-built Google ADK agent.
1. Set up your API Key:
Navigate to rabbithole/agent/adk/, copy .env.template to .env, and add your Google API key.
cd rabbithole/agent/adk
cp .env.template .env
# Now, edit .env to add your key2. Launch the Agent Server:
From the rabbithole/agent/adk directory, run:
python __main__.pyThe agent is now live on http://localhost:10000.
3. Talk to Your Agent (from a new terminal): Navigate to the CLI directory and start the client.
# In a new terminal, from the project root
cd ../../cli
python __main__.py --agent http://localhost:10000You're now chatting with your first agent! For more detailed instructions, check out How to Run and Test.
Our documentation provides a deep dive into the framework's architecture, vision, and usage.
- Live Documentation Website: The best place to start, with full search and navigation.
- Getting Started: Your main guide to installation, core concepts, and contributing.
- Framework Vision: Understand the "why" behind RabbitHole and its strategic direction.
- Software Architecture: A technical deep-dive into the system's design.
- Feature Ideas: See the list of potential upcoming features.
- Official Roadmap: The planned milestones for the project.
- Developer Guide: Everything you need to know to contribute code.
Our vision is to build the premier open-source framework for creating and orchestrating multi-agent systems. To understand our goals and where the project is headed, please see our detailed documents:
RabbitHole is an open-source project, and we welcome contributions from the community! Whether it's adding a new agent integration, improving documentation, or fixing a bug, your help is valued.
Please read our Developer Guide to get started.
This project is licensed under the MIT License.