Agentic AI Foundation Adds Four Agent Standards to Reach 10 Projects

The Agentic AI Foundation has added four new projects to its roster, taking the young standards body from three launch projects to ten hosted projects in its first year. According to the foundation, the additions were announced on October 22, 2026, at AGNTCon + MCPCon North America in San Jose, and each one was contributed by a different company: Agent Skills from Anthropic, Open Responses from OpenAI, OpenEnv from Hugging Face, and OpenSandbox from Alibaba.

The Agentic AI Foundation, known as AAIF, was formed under the Linux Foundation in December 2025. Its stated job is to give agentic AI a neutral home, so that no single company controls the shared standards, protocols, and infrastructure that agents rely on. The four new projects join a roster that already includes the Agent2Agent protocol, Agent Router, agentgateway, AGENTS.md, goose, and the Model Context Protocol, reported by the Linux Foundation in its announcement.

Read together, the four additions fill in layers of the agent stack that previously had no neutral home. That is the more useful way to read the news: individually each project is modest, but collectively they cover how agents gain new skills, how applications talk to models across providers, how agents are trained and evaluated, and where agents run safely.

What each of the four new projects does

Agent Skills, contributed by Anthropic, is an open format for packaging instructions, scripts, and resources that agents load on demand. Anthropic launched the format a year earlier, and contributing it to the Agentic AI Foundation means rival agent builders can adopt the same packaging without depending on a single vendor's roadmap.

Open Responses, contributed by OpenAI, is an open specification for interoperable language-model interfaces, based on the OpenAI Responses API, with a shared schema and streaming events. If it gains traction, applications could switch model providers without rewriting the interface layer, which is one of the oldest lock-in problems in applied AI.

OpenEnv, contributed by Hugging Face, is a framework and community hub for interoperable reinforcement-learning environments, exposed through the Model Context Protocol and Gym. Training and evaluation environments are where agents learn to act reliably, and a shared hub could make benchmark results easier to compare across labs.

OpenSandbox, contributed by Alibaba, is a general-purpose sandbox platform with multi-language software development kits, Docker and Kubernetes runtimes, and an extensible sandbox protocol. Sandboxes matter because agents that write and execute code need a hard boundary around what that code can reach, a theme echoed across the industry's recent agent-infrastructure launches, including FireMon's agentic AI for firewall policy work and the enterprise agent rollouts covered in Sprinklr's marketing AI agents open beta.

Why neutral governance matters for agents

According to Manik Surtani, chief technology officer and co-founder of the Agentic AI Foundation, the growth from three projects to ten in a year shows that companies are willing to hand shared infrastructure to neutral governance. In the foundation's announcement, Surtani said the four projects each cover a different part of the stack, from how agents gain capabilities to where they run, and that four different companies chose to contribute them to neutral governance.

That willingness is the real signal. Agent standards only work if competitors adopt them, and competitors rarely adopt a standard owned by a rival. A Linux Foundation home gives the Agentic AI Foundation a governance model the industry already trusts, with transparent processes and no single company controlling the outcome, as reported by Unite.AI in its coverage of the announcement.

The timing also matters. The announcement coincided with AGNTCon + MCPCon North America, the open agent ecosystem's flagship conference, where implementation experience and shared approaches to building, securing, and operating agents were on the agenda. The Agentic AI Foundation now hosts ten projects spanning agent-to-agent communication, routing, gateways, skills, model interfaces, training environments, and sandboxes, which is close to a full stack for building agents that can work across organizational boundaries.

What to watch next

The question now is adoption. Standards announcements are cheap; implementations are the test. Watch for agent frameworks and enterprise platforms shipping support for Agent Skills packaging, Open Responses interfaces, OpenEnv training hubs, and OpenSandbox runtimes in their default configurations, rather than as optional extras.

Watch, too, for whether the Agentic AI Foundation keeps attracting contributions from companies outside the original circle. Its first year brought in the Model Context Protocol from Anthropic and now four more projects from four more contributors. If that pace continues, the foundation could become the default venue where agent interoperability gets decided, which would be a significant shift for an ecosystem that has spent the last two years building in parallel silos.

For agent builders, the practical takeaway is straightforward: build against the open specifications where they exist, and treat vendor-specific agent interfaces as a convenience rather than a foundation. The Agentic AI Foundation is betting that the agent economy will reward interoperability, and with ten projects now under neutral governance, that bet has more weight behind it than it did a year ago.

Sources: Agentic AI Foundation announcement, Linux Foundation press release, and Unite.AI coverage.