lupantech/AgentFlow
AgentFlow is a framework for optimizing agentic AI systems using reinforcement learning to improve multi-agent coordination and tool use.
Not currently ranked — collecting fresh signals.
star history
AgentFlow provides a system for optimizing in-the-flow agentic AI, meaning it trains and refines agents during their operational lifecycle rather than just at design time. The framework supports multi-agent systems with LLM-based reasoning and integrates tool-augmented capabilities. It includes a Gradio demo, published research on arXiv, and model weights on Hugging Face.
Frequently asked
- What is lupantech/AgentFlow?
- AgentFlow is a framework for optimizing agentic AI systems using reinforcement learning to improve multi-agent coordination and tool use.
- Is AgentFlow open source?
- Yes — lupantech/AgentFlow is open source, released under the MIT license.
- What language is AgentFlow written in?
- lupantech/AgentFlow is primarily written in Python.
- How popular is AgentFlow?
- lupantech/AgentFlow has 2k stars on GitHub.
- Where can I find AgentFlow?
- lupantech/AgentFlow is on GitHub at https://github.com/lupantech/AgentFlow.