Agentic AI Foundation Expands to 10 Projects With Donations From OpenAI, Anthropic, Hugging Face

Major Tech Companies Donate AI Agent-Related Projects to Vendor-Neutral Organization

The Agentic AI Foundation has expanded its portfolio of open-source AI tools by accepting four new projects from OpenAI, Anthropic, Hugging Face, and Alibaba1

1

2

2

. This brings the total number of projects under AAIF stewardship to 10, up from three when the Linux Foundation launched the vendor-neutral organization in December 20251

1

. The donated technologies—OpenSandbox, Agent Skills, OpenEnv, and Open Responses—provide essential functions for AI agent development that developers building AI agents may find helpful1

1

.

Open Responses Tackles API Fragmentation Challenge

OpenAI contributed Open Responses, an open API specification designed to address a critical challenge facing developers working with LLM providers1

1

. The project provides a standardized template for AI model APIs that enables developers to move code between different providers without making extensive changes2

2

. Open Responses standardizes common request and response formats, streaming events, and tool invocation patterns, building LLM interfaces that can work with different model providers1

1

. Since OpenAI modeled the specification after its own API, workloads can be easily moved between OpenAI and APIs based on Open Responses, while many developers' existing familiarity with OpenAI's APIs should ease adoption2

2

.

Anthropic Donates Agent Skills for Enhanced Task Guidance

Anthropiccontributed Agent Skills, which outlines a standardized format for AI agent skills that guide AI agents through specified tasks more effectively1

1

. The format provides text descriptions and context that helps agents accomplish their goals by including instructions for carrying out tasks, illustrative examples, and automation scripts1

1

2

2

. Packaging skills in this standardized format makes sharing between development teams easier, as there's no need to restructure files before connecting them to workloads2

2

. Organizing data consistently also makes it easier to process for AI agents, which lowers the risk of hallucinations2

2

. "We're committed to keeping Agent Skills open, community-driven, and vendor-neutral," Anthropic stated1

1

.

Sandboxing Tools From Hugging Face and Alibaba

Hugging Face and Alibaba donated OpenEnv and OpenSandbox respectively, both focused on managing sandboxes used in AI training2

2

. OpenEnv offers a containerized service for creating reinforcement learning environments that can be built without much trouble, can interoperate, and can scale to meet demanding testing requirements1

1

. The technology reduces the amount of work involved in switching training systems2

2

. OpenSandbox provides a way to constrain or sandbox the capabilities of AI agents using various client SDKs, tools, protocols, lifecycle hooks, container runtimes, and assorted in-sandbox execution components1

1

. It provides a set of APIs for managing sandboxes and eases tasks such as connecting agents that run inside sandboxes to external applications2

2

.

Rapid Growth Signals Industry Confidence in Open-Source Agentic Tooling

The AAIF's nurturing of these technical specifications has attracted considerable interest from companies convinced there's a future in AI agents1

1

. The organization has surpassed 300 members in less than a year, with 60 companies joining this quarter alone1

1

. New Gold Members include Ant Group, Gates Foundation, MongoDB, Nutanix, Royal Bank of Canada, Vercel, and Vultr1

1

. Anthropic engineer David Soria Parra emphasized the foundation's importance: "The AAIF has done important work in giving the Model Context Protocol (MCP) a neutral home since we donated it less than a year ago, and alongside the AAIF, we're committed to continuing to invest in open source agentic tooling for builders everywhere"2

2

. However, this rapid embrace of AI agents may be moving too quickly for some sectors, as US banks have raised concerns that AI agents are interacting with customer accounts without bank authorization1

1

.