“Be ready for change”: AAIF’s CTO on why the agent stack won’t settle

The New Stack (AI)
Read full story· 6 min
The Agentic AI Foundation (AAIF) expanded its open infrastructure for AI agents by adding four major projects from Anthropic, OpenAI, Hugging Face, and Alibaba, aiming to standardize agent communication and reduce vendor lock-in.

What happened

On October 22, the Agentic AI Foundation (AAIF) announced that four major projects from Anthropic, OpenAI, Hugging Face, and Alibaba had joined its open infrastructure for AI agents. This expansion brought the total number of projects hosted by AAIF to ten. The announcement was made by AAIF’s CTO, Manik Surtani, during his keynote at AGNTCon + MCPCon in San Jose. The foundation, backed by the Linux Foundation, aims to provide a vendor-neutral home for shared agent infrastructure and standards.

Key details

  • AAIF was launched in December 2025 with three initial projects, including the Model Context Protocol (MCP).
  • The four new projects are Agent Skills (Anthropic), Open Responses (OpenAI), OpenEnv (Hugging Face), and OpenSandbox (Alibaba).
  • AAIF introduced a new Sandbox lifecycle phase on September 1 to support early-stage projects.
  • AAIF has over 300 member organizations within its first 10 months.
  • Agent Skills and Open Responses are open formats and specifications for agent instructions and model communication.

Why it matters

Software teams and organizations building AI agents now have access to more standardized, vendor-neutral tools for agent communication, task execution, and environment management. This reduces the risk of being locked into a single provider and makes it easier to switch or integrate different agent technologies. Developers can adopt these open standards to improve portability and collaboration, while early-stage projects can join the foundation’s Sandbox phase for neutral governance and support.


More in Agents