Stars
This repository contains the official implementation of "FastVLM: Efficient Vision Encoding for Vision Language Models" - CVPR 2025
🔥 Open Source Browser API for AI Agents & Apps. Steel Browser is a batteries-included browser instance that lets you automate the web without worrying about infrastructure.
🦉 OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation
MarinaBox is a toolkit for creating and managing secure, isolated environments for AI agents
Atropos is a Language Model Reinforcement Learning Environments framework for collecting and evaluating LLM trajectories through diverse environments
Open source interpretability artefacts for R1.
Official Repo for InSTA: Towards Internet-Scale Training For Agents
HumanLayer enables AI agents to communicate with humans in tool-based and async workflows. Guarantee human oversight of high-stakes function calls with approval workflows across slack, email and mo…
🦄 ai that works - every tuesday 10 AM PST
What are the principles we can use to build LLM-powered software that is actually good enough to put in the hands of production customers?
Pack your repo into an AI-friendly format and apply AI suggested changes back with ease, all from your terminal.
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflo…
Lightweight coding agent that runs in your terminal
Agent Reinforcement Trainer for training multi-turn agents using GRPO
Challenges for general-purpose web-browsing AI agents
A mini, open-weights, version of our Proxy assistant.
SoTA production-ready AI retrieval system. Agentic Retrieval-Augmented Generation (RAG) with a RESTful API.
🌎💪 BrowserGym, a Gym environment for web task automation
An AI web browsing framework focused on simplicity and extensibility.
The SOTA Open-Source Browser Agent for autonomously performing complex tasks on the web
Build your own visual reasoning model
An open protocol enabling communication and interoperability between opaque agentic applications.
Code for Paper: Training Software Engineering Agents and Verifiers with SWE-Gym [ICML 2025]
Eval whether LLMs can win small Perfect Information Games (PIG) from position where perfect play is a win
An agent benchmark with tasks in a simulated software company.
Computer Agent Arena: Test & compare AI agents in real desktop apps & web environments. Code/data coming soon!