Agentic AI Foundation Announces Four New Projects: Agent Skills, Open Responses, OpenEnv, and OpenSandbox

Agentic AI Foundation Announces Four New Projects: Agent Skills, Open Responses, OpenEnv, and OpenSandbox

PR Newswire

SAN JOSE, Calif., Oct. 22, 2026

AAIF reaches 10 projects in its first year, with new contributions from Alibaba, Anthropic, Hugging Face, and OpenAI

Summary

• AAIF announced four new projects: Agent Skills, contributed by Anthropic; Open Responses, contributed by OpenAI; OpenEnv, contributed by Hugging Face; and OpenSandbox, contributed by Alibaba.
• AAIF grew from three projects to 10 in its first year.
• The new projects were announced at AGNTCon + MCPCon North America. They expand AAIF's work in AI agent skills, connections between different AI models, training and testing environments, and safe places to run agents. All are managed under open rules that do not favor any one company.

SAN JOSE, Calif., Oct. 22, 2026 /PRNewswire/ -- AGNTCon + MCPCon North America — The Agentic AI Foundation (AAIF) announced four new projects today: Agent Skills, contributed by Anthropic; Open Responses, contributed by OpenAI; OpenEnv, contributed by Hugging Face; and OpenSandbox, contributed by Alibaba. These additions bring the number of projects hosted by AAIF to 10 in its first year, up from the three projects it launched with in December 2025.

The new projects cover four parts of the AI agent system: how agents learn new skills, how applications communicate with AI models from different providers, how agents are trained and tested, and where agents can run safely. They join AAIF projects including the Agent2Agent (A2A) protocol, Agent Router, agentgateway, AGENTS.md, goose, and the Model Context Protocol (MCP). All these projects follow AAIF's open rules, which are designed to prevent any one company from controlling the work.

"One year ago, AAIF started with three projects and the belief that AI agents needed a neutral place to grow. Today, we host 10," said Manik Surtani, CTO and co-founder of the Agentic AI Foundation. "Agent Skills, Open Responses, OpenEnv, and OpenSandbox each cover a different part of the system, from helping agents gain new abilities to providing places for them to run. Four different companies chose to contribute these projects to a neutral organization. That is what this foundation is for: giving the community one place to build shared standards, protocols, and infrastructure, without letting any single company control the outcome and while putting the community first."

The announcement comes alongside AGNTCon + MCPCon North America, taking place October 22-23 at the San Jose McEnery Convention Center. The conference brings together people working on open AI agents to share practical experience, discuss common problems, and improve ways to build, protect, and run AI agents.

New Projects

Agent Skills is an open format that helps AI agents gain new abilities and knowledge about specific fields. It bundles instructions, scripts, and resources in a reusable package. Agents can load these skills when needed, making specialized knowledge usable across different agent products and platforms.

"Even the smartest agents can fall short when they do not have the right information and context. We launched Agent Skills a year ago to give developers and businesses an open standard that makes their agents more capable. Today, we are donating Agent Skills to AAIF so that everyone building with skills can help decide what comes next. AAIF has done important work by giving the Model Context Protocol (MCP) a neutral home since we donated it less than a year ago. We will continue working with AAIF to support open-source tools for people building AI agents everywhere." – David Soria Parra, Co-Creator and Lead Maintainer of MCP, Anthropic

Open Responses is an open set of rules for building AI language model interfaces that work with different providers. It is based on the OpenAI Responses API. It defines a shared data format, consistent events for sending results as they are created, and tools that can be expanded. This lets developers describe requests and results once and use them with different model providers with very little change.

"Open Responses is excited to join AAIF as we continue working to help developers build great products using today's language models through simple, clear, and unified interfaces. AAIF has a strong record of supporting important projects across the industry, and we are excited to keep expanding what developers can achieve." – Steve Coffey, Member of Technical Staff, OpenAI

OpenEnv is an open framework and community hub for creating reinforcement learning (RL) environments. These environments help AI agents learn through practice. OpenEnv is designed to make them work together, easy to connect and use, simple to build and share, and able to grow to support more work. It uses a simple service that runs in a container and provides a familiar interface for agents, the Model Context Protocol (MCP), and Gym, the protocol used by training libraries.

"We believe more than ever that making AI available to everyone requires a fully open set of tools and systems. Reinforcement learning environments are a key part of this system because they teach agents the abilities they will need in the real world. The committee behind OpenEnv is proud to bring AAIF into reinforcement learning by sharing the work we have built and used at Hugging Face, Microsoft, Reflection, Meta-PyTorch, Fleet, Mercor, RadixArk, Modal, Prime Intellect, and other organizations." – Ben Burtenshaw, Machine Learning Engineer, Hugging Face

OpenSandbox is a general-purpose sandbox platform for AI applications. It provides software development kits (SDKs) for multiple programming languages, a shared set of sandbox APIs, and Docker and Kubernetes systems for running software. It supports uses such as coding agents, graphical user interface (GUI) agents, agent testing, AI code execution, and reinforcement learning training. It also includes a sandbox protocol that defines how sandboxes are started, managed, and used to run tasks. This allows teams to add their own systems for running software.

"Although OpenSandbox was created at Alibaba, we do not want one company to determine its future. By bringing the project to AAIF, we are making it possible for more developers and organizations to help shape how sandbox environments work across the agent ecosystem." – Xiaobin Xu, Head of Development Infrastructure, Alibaba

About the Agentic AI Foundation
The Agentic AI Foundation (AAIF) is a neutral home for open standards, protocols, and infrastructure used in AI agent systems. Its founding projects include MCP, goose, and AGENTS.md. AAIF oversees the main standards, protocols, and frameworks that help agents work together across different platforms. Through open decision-making and participation from many parts of the industry, AAIF works to encourage the use of these tools and ensure that AI agent infrastructure develops openly, fairly, and at a scale suitable for real-world use. For more information, please visit aaif.io.

Media Contact
Sarah Zonouzi
Agentic AI Foundation PR
[email protected]

View the original content and download multimedia: https://www.prnewswire.com/news-releases/agentic-ai-foundation-announces-four-new-projects-agent-skills-open-responses-openenv-and-opensandbox-302916689.html

SOURCE Agentic AI Foundation