---
slug: "reef"
name: "Reef"
effect: "Turns feedback on an agent's live requests into versioned updates to its model weights or its harness"
category: "agent-framework"
tier: "rising"
url: "https://reefinfra.ai"
github: "Human-Agent-Society/reef"
pricing: "open-source"
license: "Apache-2.0"
tags: ["self-improving-agents", "training", "harness-optimization", "feedback", "inference-server"]
platforms: []
install_kind: "pip"
date_added: "2026-10-15"
rank: 2
rising_score: 74
stars: 7602
star_velocity: 168.4
velocity_basis: "since creation"
created_at: "2026-08-31T01:39:18Z"
pushed_at: "2026-10-15T04:56:46Z"
signals_status: "ok"
signals_fetched_at: "2026-10-15T05:02:34.593Z"
page: "https://agenttooldrop.com/tools/reef"
---

# Reef

Turns feedback on an agent's live requests into versioned updates to its model weights or its harness

Category: Agent frameworks. Tier: new and rising, rank 2 by new and rising score (74 of 100). Pricing: Open source. License: Apache-2.0.

## Why you would want it

Reef sits between an agent and its model as an OpenAI- and Anthropic-compatible inference endpoint, records each interaction and returns a receipt that a later report can score or comment on. Recipes turn eligible feedback into updates, either training model weights with Slime and SGLang on a GPU stack or refining the harness, meaning prompts, rules and skills, through a model API with no local GPUs. Accepted updates go into a version history and are served without restarting Reef. The built-in Reefine recipe refines a coding harness from plain-language asks, such as reproducing a bug with a failing test first.

## Best for

- Agents that should learn from user feedback
- Harness optimization without GPUs
- Continual learning research

## Use it when

You run an agent with a steady stream of tasks or feedback and want that signal to produce evaluated, versioned improvements instead of manual prompt edits.

## Skip it if

You want a stable agent today: harness work needs a model endpoint, representative tasks and an evaluator, weight training needs a supported GPU stack, and it is at v0.1.3 with the repo created on 2026-08-31. The harness install pipes a script into bash and the pi adapter runs commands with no sandbox, so keep the install root outside the project the agent works in.

## Editor note

Benchmark results in the recipe pages are the project's own; we do not repeat them.

## Install

Install from PyPI with uv:

```sh
uv venv && source .venv/bin/activate
uv pip install reef-infra
python3 -c "import reef; print(reef.__version__)"
```

Source: https://github.com/Human-Agent-Society/reef#from-pypi

## Live signals

Fetched 2026-10-15T05:02:34.593Z (status: ok).

- Repository: Human-Agent-Society/reef
- Repository created: 2026-08-31
- Stars: 7,602 (GitHub's count; it cannot be authenticated)
- Stars per day, since repo creation: 168.4 (an average over the repo's life, not a growth rate)
- Forks: 709
- Open issues and PRs: 91
- License (GitHub): Apache-2.0
- Last push: 2026-10-15
- Latest release: v0.1.3 (2026-10-14)

## Links

- Website: https://reefinfra.ai
- GitHub: https://github.com/Human-Agent-Society/reef
- Install instructions verified at: https://github.com/Human-Agent-Society/reef#from-pypi
- Source for this entry: https://github.com/Human-Agent-Society/reef#readme
- AgentToolDrop page: https://agenttooldrop.com/tools/reef
- Full catalog (JSON): https://agenttooldrop.com/tools.json