An awesome list that summarizes open-source repositories for training LLM agents using reinforcement learning, categorized into base frameworks, general/multi-task, search/research/web, GUI, and tool-based agents.
One-Minute Overview#
AgentsMeetRL is a curated list of open-source projects focused on training Large Language Model (LLM) agents using reinforcement learning. It collects the latest research projects from academia and industry across multiple domains including base frameworks, multi-task learning, intelligent search, GUI interactions, and tool usage.
Core Value: Provides researchers and developers with a structured reference for reinforcement learning agent implementations, helping understand different technical approaches and solutions.
Quick Start#
Installation Difficulty: Low - This is a project compilation list that requires no installation. Simply access the information directly.
Is this suitable for my needs?
- ✅ Researchers: Looking for reinforcement learning agent implementations and latest research findings
- ✅ Developers: Seeking reusable reinforcement learning frameworks and algorithm implementations
- ❌ End users wanting ready-to-use products: This is merely a resource list, not a runnable software product
Core Capabilities#
1. Categorized Project Display - Solves Information Overload#
- Projects classified into five main categories: base frameworks, general/multi-task, search/research/web, GUI, and tools User Value: Helps users quickly identify project types that match their specific needs
2. Detailed Technical Comparison - Solves Selection Challenges#
- Each project includes key information such as RL algorithms, single/multi-agent setups, reward mechanisms, and single/multi-turn capabilities User Value: Enables developers to compare technical approaches across different projects for more informed decision-making
3. Real-time Update Tracking - Solves Timeliness Issues#
- Each project is labeled with release date and institution, ensuring information is current and authoritative User Value: Helps users stay current with research frontiers and avoid outdated technical solutions
Technology Stack & Integration#
Development Languages: Various languages used, including Python, JavaScript, etc. Main Dependencies: Built on mainstream deep learning frameworks like PyTorch, HuggingFace Transformers, etc. Integration Methods: API / SDK / Library - Most projects provide complete training frameworks or libraries
Maintenance Status#
- Development Activity: Very active, with continuous updates and numerous new projects added in 2025
- Recent Updates: Frequently updated to stay synchronized with the latest research
- Community Response: Open contribution mechanism allows users to submit new projects or corrections via PRs and Issues
Documentation & Learning Resources#
- Documentation Quality: Comprehensive - Each project provides detailed technical information and reference links
- Official Documentation: Accessible through GitHub repositories for each project
- Sample Code: Most projects provide complete code implementations and examples