ART

Open-source reinforcement learning framework that trains LLM agents on multi-step tasks using GRPO.

IndexedApache-2.0★ 10,819

Install

From the package registry. Read the repository before you give any tool access to your data.

pip install openpipe-art

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

About

ART (Agent Reinforcement Trainer) is an open-source RL framework that improves agent reliability by letting LLMs learn from experience. It gives Python applications a harness for integrating GRPO training, with example notebooks for tasks such as email search, playing 2048 and an MCP server. W&B Training offers a serverless RL service that manages the training and inference infrastructure.

Use cases: model-inference

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Fact sheet

Languages: PythonREADME
Packages: openpipe-art (pypi)README
Deploys to: wandb-inferenceREADME
Maturity: Betagithub.com

Facts extracted by Claude Haiku 5.5 from the linked sources on Oct 24, 2026; reviewed Oct 24, 2026. Wrong fact? Tell us.

Alternatives

Frameworks for the same language with a similar way of orchestrating agents.

Compare and alternatives

Signals

Stars: 10,819
Forks: 1,009
Open issues: 83
Weekly downloads: 2,536
Last commit: Oct 22, 2026 (4 days ago)
Last release: Oct 20, 2026 (7 days ago)
Last package publish: Oct 20, 2026 (7 days ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Packages: pypi openpipe-art
License: Apache-2.0
Repository: github.com/OpenPipe/ART
Website: art.openpipe.ai
First seen: Oct 24, 2026
Last checked: Oct 24, 2026
Source: README github-graphql

Cite this page

ART. Agent Mart. https://agentmart.dev/harnesses/art (last updated 2026-10-24)