Find tools › Open-source projects › Write code faster

reef

Open-source project · by Human-Agent-Society

Infrastructure for AI agents that automatically learns and improves from user feedback and interactions over time.

Some setupWell known1,628 installs a month

What it costs

Free and open source; you may pay for the AI model it uses.

Use it to

  • Train model weights using runtime interaction feedback
  • Refine agent prompts, rules, and skills automatically
  • Serve OpenAI-compatible agent requests with version management

How to start

  1. Create and activate a virtual environment using uv, then install the package.

  2. Start Reef as a pure inference server using uv run reef serve --inference.model-path Qwen/Qwen2.5-1.5B-Instruct.

  3. Send inference requests to the endpoint and report feedback to trigger learning cycles.

Good to know. Requires Git LFS for artifact tracking and a suitable GPU setup for weight-training configurations. The repository was pushed recently, but model training workloads carry inherent infrastructure complexity.

Before you install

  • Found automaticallyNot reviewed by a person yet. Check it before installing.
  • License: Apache-2.0You are allowed to use it under these terms.
  • Last updated 2026-11-18

More detail

Reef is an open-source infrastructure platform that connects agent inference, feedback, learning, and versioned delivery to continually improve AI models or harnesses.

It solves the problem of manually updating and training agents by automatically processing interaction feedback into weight updates or harness improvements. Reach for it when building self-improving agent systems.

Similar tools

Descriptions are written in plain English from each project's own documentation and refreshed by our agent; this one was last checked 2026-09-29. Spotted a mistake? · Open in Find tools · © 2026 Which AI Map