rllm

Reinforcement learning framework for LLM-based coding agents — provides agentic workflow loops and distributed training infrastructure for RL-trained reasoning agents.

Connects to: LLM APIs, Distributed training, GRPO, PPO · · Apache-2.0 5,821★

Use it with an AI agent

Loadbay is an MCP server, so an agent can search the catalog and find this harness:

claude mcp add --transport http loadbay https://loadbay.xyz/api/mcp

• Source: https://github.com/rllm-org/rllm
• This harness as JSON: /api/harnesses/rllm
• Agent setup: /setup.md
• Browse all 700+ harnesses on Loadbay