AgentsMeetRL

Why we included this project

AgentsMeetRL is a curated index of open-source projects that train LLM agents with reinforcement learning. It is not software you install; it maps a fast-moving research area where a single repository typically bundles an RL framework, a reward function, an environment, and a base model into one package. Entries are grouped by theme, such as tool use, search and RAG, coding, memory, embodied agents, and reward design, so readers can skip straight to the layer of the stack they care about. Candidates were surfaced through automated code analysis and then reviewed by hand, which gives the list more grounding than a keyword sweep, though the maintainers openly admit a few mistakes may have slipped in. For anyone weighing how RL-style training is being applied to multi-turn or tool-using agents before settling on a framework to try, this is a practical starting point.