TRL plus OpenEnv can now train agents inside custom harnesses like OpenCode and Claude Code
ben_burtenshaw · x · 2026-07-24
You can now train agents in TRL and OpenEnv with a harness of your choice, including OpenCode, pi, Codex, or Claude Code.
This is a training setup for agentic workflows, not just a prompt trick:
- Start from a dataset of tasks.
- Use a harness such as OpenCode to drive the session.
- Connect an OpenEnv environment and a harness adapter.
- Collect traces with TRL HarnessRolloutWorker.
- Use AsyncGRPOTrainer to orchestrate inference and weight updates, with the verifier returning envreward.
The attached diagram shows a full loop: task + verifier, loop-owning agent, environment execution, rollout collection, and policy refresh.
Related event: TRL and OpenEnv Enable Custom Agent Harness Training(4 posts)→
More from coding & agent
- Dev turns DJI Mavic into local AI security drone built with Opus 5.5 — alexcovo_eth · 2026-10-27
- Microsoft Research: ForkPilot cuts long-horizon agent compute by up to 59% — EliasEskin · 2026-10-27
- Cursor launches Agent SDK: agents as code with Markdown prompts and TypeScript tools — soleio · 2026-10-27
- Vibecheck, the 'usable formal methods' hackathon, sells out two floors in SF — spikedoanz · 2026-10-27
- How do you verify AI agent outputs beyond unit tests and LLM-as-judge? — Practical_Green5775 · 2026-10-27
- AI agents may soon become the only way to run code, not just write it — IgorCarron · 2026-10-27