Agentic AI Foundation Reaches 10 Projects: Anthropic, OpenAI, Alibaba Unite for Open Agent Standards

Introduction

On October 22, 2026, at the AGNTCon + MCPCon North America conference in San Jose, California, the Agentic AI Foundation (AAIF) announced the addition of four major open-source projects, bringing its total hosted projects to 10. These four projects were contributed by Anthropic, OpenAI, Hugging Face, and Alibaba, marking the first time that the world's leading AI companies have collaborated under a neutral governance framework to define infrastructure standards for agentic AI.

Background and Mission of AAIF

The Agentic AI Foundation was established in December 2025 under the Linux Foundation, initially with just three projects. Its core mission is to provide a neutral, open home for agentic AI technology, ensuring that no single company can dominate the evolution of the agentic AI ecosystem.

In less than a year, AAIF has expanded from three to ten projects, covering all critical layers of the agentic AI stack, including existing projects such as the Model Context Protocol (MCP), Agent2Agent (A2A), Agent Router, agentgateway, AGENTS.md, and goose.

The Four New Projects in Detail

1. Agent Skills (Contributed by Anthropic)

Agent Skills is an open-source format for packaging task-specific instructions, scripts, and resources. Using a SKILL.md file, this format enables agents to load specialized capabilities on demand.

Anthropic reported that the standard is supported by over 70 agent platforms and has seen widespread adoption. This means developers can create reusable "skill packages" that allow different AI agents to share and invoke the same specialized capabilities, significantly reducing the cost of redundant development.

2. Open Responses (Contributed by OpenAI)

Open Responses is a specification for multi-provider, interoperable LLM interfaces based on the OpenAI Responses API. It provides a shared schema and consistent streaming events, allowing developers to build against a single interface that functions across various model providers with minimal translation requirements.

This contribution is highly significant—OpenAI opening its core API specification to the entire industry means developers no longer need to write different integration code for each model provider, dramatically reducing the complexity of multi-model application development.

3. OpenEnv (Contributed by Hugging Face)

OpenEnv is a framework and community hub for creating interoperable reinforcement learning (RL) environments. It uses containerized services that support both the Model Context Protocol (MCP) and the Gym training protocol.

For researchers and developers who need to train agents to perform tasks in complex environments, OpenEnv provides a standardized training environment that makes it easier to compare and reproduce research results across different institutions.

4. OpenSandbox (Contributed by Alibaba)

OpenSandbox is a general-purpose, self-hosted sandbox platform for AI applications. It offers multi-language SDKs and unified APIs with runtimes for Docker and Kubernetes, facilitating secure execution for coding agents, agent evaluation, and AI-driven task execution.

Notably, OpenSandbox recorded over 1,400 commits in the 90-day period leading up to the announcement, demonstrating extremely high development activity. Alibaba's contribution of this core infrastructure to a neutral foundation signals Chinese tech giants' active participation in global AI standard-setting.

Governance Architecture and Openness

AAIF's governance model emphasizes multi-party participation and transparency. All new projects will establish Technical Steering Committees and formal proposal processes, recruiting maintainers from multiple organizations to ensure ongoing development and security.

This governance model differs from traditional enterprise-led open-source projects—under the AAIF framework, even the companies that contributed projects cannot unilaterally determine the project's direction; all major decisions require community consensus.

Impact on the Asia-Pacific Region

For AI developers and enterprises in the Asia-Pacific region, AAIF's expansion has important implications:

Lowering Technical Barriers: Standardized agent interfaces and skill formats mean that Asia-Pacific developers can more easily build AI agent applications compatible with the global ecosystem.

Alibaba's Participation: Alibaba's contribution of OpenSandbox is not just a technical contribution but also a signal of Chinese tech enterprises playing an active role in global AI governance. This serves as a model for Asia-Pacific AI companies seeking to compete in global markets.

Accelerating Enterprise Adoption: As agentic AI standards gradually solidify, Asia-Pacific enterprises can invest in agentic AI technology with greater confidence, without worrying about vendor lock-in.

Industry Significance and Future Outlook

AAIF's expansion to 10 projects in less than a year reflects the entire industry's urgent need for agentic AI standardization. One of the biggest challenges facing the agentic AI market today is fragmentation—agents on different platforms cannot communicate with each other, and tools from different vendors are difficult to integrate.

AAIF's efforts are changing this situation. By establishing common protocols, interfaces, and execution environments, AAIF is laying the foundation for a more open, interoperable agentic AI ecosystem.

Analysts predict that as these standards mature and are adopted, the cost of deploying agentic AI will decrease significantly, and the time for enterprises to build complex multi-agent systems will be reduced from months to weeks.

Conclusion

The Agentic AI Foundation's expansion to 10 projects is an important milestone in the maturation of the agentic AI industry. The collaboration of Anthropic, OpenAI, Hugging Face, and Alibaba under a neutral governance framework demonstrates that even fiercely competing companies can find common ground on infrastructure standards.

For AI practitioners in the Asia-Pacific region, now is the best time to deeply understand and participate in these open standards—they will profoundly shape the development landscape of agentic AI in the coming years.