The foundation now hosts ten projects. The Agentic AI Foundation announced four new hosted projects on October 22, 2026, bringing its total to 10 in its first year after launching with three projects in December 2025.
“A year ago, AAIF launched with three projects and a bet that agentic AI needed a neutral home to thrive. Today we host 10,” said Manik Surtani, CTO and co-founder of the Agentic AI Foundation. The new additions came from Anthropic, OpenAI, Hugging Face, and Alibaba.
Four projects cover the agent stack
Anthropic contributed Agent Skills, OpenAI contributed Open Responses, Hugging Face contributed OpenEnv, and Alibaba contributed OpenSandbox. Together, the projects address how agents gain new skills, how applications communicate with models across providers, how agents are trained and evaluated, and where agents run safely.
Agent Skills is an open format for giving agents new capabilities and domain expertise. Its packages combine instructions, scripts, and resources in a reusable form that agents can load on demand, allowing the same skill format to serve different tasks without rebuilding each capability from scratch.
More than 70 agent platforms already support Agent Skills, while over 1 million skills have been published with nearly 280 million installations. The skills cover software engineering, marketing, design, writing, and research — a broad list of uses for a format that is still in its first year of AAIF hosting.
The skills.sh registry listed 1 million skills in seven months. That figure describes the registry’s listings, while the broader Agent Skills figures cover published skills and installations, giving the project three separate measures of adoption instead of one convenient headline.
Open Responses is an open specification for building multi-provider, interoperable LLM interfaces, based on the OpenAI Responses API. It defines a shared schema, consistent streaming events, and extensible tooling for requests and outputs across model providers.
Steve Coffey, a member of technical staff at OpenAI, said the project is joining AAIF to help developers build products on language models through simple, unified interfaces. The aim is a common way for applications to work with models from different providers, because rebuilding the same interface for every provider is a fine use of time only if nobody values time.
AAIF adds infrastructure around existing projects
The four projects join existing AAIF projects including the Agent2Agent, or A2A, protocol; Agent Router; agentgateway; AGENTS.md; goose; and the Model Context Protocol, or MCP. David Soria Parra is a co-creator and lead maintainer of MCP at Anthropic.
That list gives AAIF projects across several parts of agent development, from protocols and routing to context, gateways, skills, model interfaces, training, evaluation, and execution environments. The new projects do not replace the launch projects; they expand the set of hosted work from three projects in December 2025 to 10 in the foundation’s first year.
OpenEnv addresses how agents are trained and evaluated, while OpenSandbox addresses where agents run safely. The announcement provides no additional project details for either one, but their inclusion gives the foundation coverage beyond the instructions and interfaces that developers use to build agent systems.
AAIF’s first-year total now includes contributions from Anthropic, OpenAI, Hugging Face, and Alibaba alongside its existing projects. The foundation’s stated model is a neutral home for agentic AI projects, and its latest additions put that model under pressure to coordinate several layers of the stack rather than merely collect names.
Based on
