Published October 23, 2026. Last updated October 23, 2026.
Four companies that fight each other for customers just agreed on something. They handed pieces of their AI agent technology to a neutral home, so nobody owns the rules.
That home is the Agentic AI Foundation, or AAIF, which sits under the Linux Foundation. The news dropped on October 22, 2026, at AGNTCon + MCPCon North America in San Jose.
If you build with AI tools, or just use them, this is worth five minutes. Here’s the plain version.
Quick answer
What happened? The AAIF announced four new projects: Agent Skills from Anthropic, Open Responses from OpenAI, OpenEnv from Hugging Face, and OpenSandbox from Alibaba.
Is the AAIF new? No. It started in December 2025 with three projects. This announcement takes it to ten. One correction to some headlines: the foundation didn’t launch this week. The four projects did.
Why should you care? Agents are the next big fight in AI. If every company builds its own incompatible way for agents to use tools, you get locked in. Shared standards make it easier to switch tools without starting over.
What is the Agentic AI Foundation?
The AAIF describes itself as the neutral home for open standards, protocols and infrastructure in agentic AI. “Agentic” just means AI that does multi-step tasks on its own, not only answers questions.
It began with three founding projects: MCP (Model Context Protocol), goose, and AGENTS.md. According to the Linux Foundation, it also hosts Agent2Agent (A2A), Agent Router and agentgateway. With the four new ones, that’s ten.
Anthropic says it co-founded the AAIF with Block and OpenAI in 2025, and that it had already donated MCP there.
The foundation’s CTO and co-founder, Manik Surtani, said the goal is to give the community “one place to build shared standards” so no single company controls them.
What are the four new AAIF projects?
1. Agent Skills (from Anthropic)
A skill is a folder with a SKILL.md file. It teaches an agent how to do one specific job, with any scripts or reference files it needs. The agent loads it only when the task calls for it.
Think of it as a how-to guide an AI can pick up on demand. Want your agent to follow your company’s report format? Write that down once as a skill.
2. Open Responses (from OpenAI)
This is an open specification for talking to different AI providers in one consistent way. It’s based on the OpenAI Responses API and defines a shared schema and consistent streaming events. The idea is that you write your app once and it works across providers.
3. OpenEnv (from Hugging Face)
OpenEnv is a framework and community hub for building reinforcement learning environments. These are the practice grounds where agents get trained and tested. It uses containerized services that expose MCP and Gym interfaces.
4. OpenSandbox (from Alibaba)
A sandbox is a locked room where an agent can run code without touching anything real. OpenSandbox is a general-purpose version of that, with multi-language SDKs, unified APIs, and Docker and Kubernetes runtimes. The Linux Foundation lists coding agents, GUI agents and RL training as use cases.
Alibaba’s Xiaobin Xu put the motive simply: “we don’t want its future to be defined by a single company.”
Why is Agent Skills the one to watch?
Because it already has traction. These numbers come from Anthropic’s own announcement, so treat them as the company’s claims:
- About 1 million skills published
- Nearly 280 million installs
- More than 70 agent platforms integrated, including ChatGPT, Cursor, GitHub Copilot and VS Code
Anthropic launched skills in October 2025 and published them as an open standard that December. Developers quickly wrote loaders for Codex, Cursor and OpenCode on their own.
With the donation, the standard moves to the AAIF along with its name, logo, GitHub organization and the agentskills.io domain. Engineers from Anthropic, Google, OpenAI and other companies will maintain it as individuals. The lead maintainers are David Soria Parra and Mahesh Murag.
Anthropic has also proposed a roadmap: versioned releases, a formal proposal process, and security guidance for people who write and install skills. You don’t need to join the AAIF to use or contribute to the standard.
What does this mean for you?
If you use AI tools every day: A skill you write for one tool can work in others that support the standard. Write your instructions once, not five times.
If you’re a developer or freelancer: Learning these standards is a safe bet. They’re open, and the biggest labs are backing them. That matters more than any single vendor’s feature list.
If you create content or run an online business: You don’t need to code. A skill is a text file. Writing down how you want a repeated task done, like your product description style or your blog checklist, is something anyone can try.
A small warning: Skills can include scripts. Install them like you’d install any software. Check what’s in a skill before you trust it. Anthropic’s own plan to publish security guidance suggests that’s a real concern.
What’s still unclear?
It’s early, so keep some doubt handy.
- The Linux Foundation’s announcement doesn’t list AAIF member companies. It names only the four donors.
- The press release doesn’t link to the project repositories.
- “Vendor-neutral” is a promise about governance. The donating companies will still have big influence early on, simply because they wrote the code. Watch who actually maintains these projects over the next year.
- The adoption numbers for Agent Skills are self-reported by Anthropic.
FAQ
What is the Agentic AI Foundation (AAIF)?
It’s a Linux Foundation project that hosts open standards and infrastructure for AI agents, including MCP, goose, AGENTS.md and now four new projects.
Which companies donated projects on October 22, 2026?
Anthropic donated Agent Skills, OpenAI donated Open Responses, Hugging Face donated OpenEnv, and Alibaba donated OpenSandbox.
What is Agent Skills?
An open format where a folder with a SKILL.md file teaches an AI agent a specific task, loaded only when needed.
What is MCP?
MCP is the Model Context Protocol, a standard for connecting AI models to tools and data. Anthropic donated it to the AAIF earlier.
Do I have to join the AAIF to use Agent Skills?
No. According to Anthropic, developers can keep using and contributing to the standard without joining the foundation.
Is the AAIF a new organization?
No. It was founded in 2025 and started with three projects in December 2025. October 22 brought four new projects, taking it to ten.
Bottom line
Rivals rarely share their best work. When they do, it’s usually because the alternative is a mess nobody wants: ten incompatible ways to do the same thing. The AAIF is a bet that agents will grow faster on shared rules than on closed ones.
Whether it stays truly neutral is something only time will show. For now, if you work with AI agents, it’s a good moment to learn what a SKILL.md file looks like.