agents
Agents 2.0 is a framework for symbolic learning that enables language agents to self-evolve by training them similarly to neural networks. It solves the problem of optimizing agent pipelines through language-based loss, gradients, and weight updates. It is intended for developers and researchers working on advanced AI agents and multi-agent systems.
✨ Key features
- Implements loss functions, back-propagation, and weight optimizer via prompt pipelines
- Supports training of single and multi-agent systems
- Provides a systematic framework for agent learning and evaluation
- Open-source with installation from git or local development
- Includes detailed documentation and paper references
🎯 Use cases
- Training language agents to improve performance on tasks
- Optimizing multi-agent systems by treating nodes as agents
- Research on self-evolving agents and symbolic learning
- Evaluating and comparing agent architectures
📦 Installation
🧰 Requirements: Python environment with pip; no specific OS or API keys mentioned.
pip install git+https://github.com/aiwaves-cn/agents@master
For local development:
git clone -b master https://github.com/aiwaves-cn/agents
cd agents
pip install -e .
❓ FAQ
What is Agent symbolic learning?
It is a systematic framework for training language agents, drawing an analogy between agent pipelines and neural networks, where prompts and tools act as weights.
How does the training process work?
It involves a forward pass (agent execution) storing a trajectory, then a prompt-based loss function evaluates the outcome, back-propagation generates language gradients, and finally all components are updated according to those gradients.
Can I use this for multi-agent systems?
Yes, the framework naturally supports optimizing multi-agent systems by considering nodes as different agents or allowing multiple agents to take actions in one node.
How do I install Agents 2.0?
You can install it directly from the git repository using pip, or clone the repository and install in editable mode for development.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.