OpenAI donates Open Responses API spec to Agentic AI Foundation

A move toward standardization

OpenAI has officially announced the donation of its Open Responses API specification to the Agentic AI Foundation (AAIF). This move is intended to create a common language for how software interacts with large language models (LLMs), which are the engines behind tools like ChatGPT.

Currently, while almost every AI company offers similar features—such as tool calling, function calling, and message handling—each company builds its own unique way for developers to access those features. This fragmentation means that if a developer wants to switch from one AI provider to another, they often have to rewrite significant portions of their code to accommodate the new provider's specific requirements.

Open Responses was designed to solve this specific headache. It provides a set of core API calls that are intended to work across different providers. By standardizing these basic building blocks, the project allows developers to maintain fewer provider-specific integrations while still having the flexibility to use different models as needed.

Joining a neutral home

By donating the project to the Agentic AI Foundation, OpenAI is placing the specification in a vendor-neutral environment. The AAIF is a part of the Linux Foundation, a well-known organization that hosts many of the world's most important open-source technologies. This structure ensures that the development of the standard is guided by a technical steering committee rather than being controlled by a single corporation.

As thenewstack.io reported, this neutral home is essential for gaining wide-scale adoption. It allows both developers and various AI providers to collaborate on the specification together. The foundation already hosts several other projects, including the MCP and A2A protocols, as well as OpenAI's AGENTS.md and the open-source AI agent goose.

The donation also includes other frameworks like OpenSandbox, Anthropic’s Agent Skills, and HuggingFace’s OpenEv framework for agent environments. Together, these projects aim to create a cohesive ecosystem for building 'agents'—AI systems that can perform multi-step tasks autonomously.

The current landscape

The industry is currently seeing a split in how these interactions are handled. While OpenAI's earlier Chat Completions API became an early standard, other companies have created their own successful versions. For example, Anthropic defines its own Messages API, which has been adopted by several other companies like DeepSeek and Moonshot.

This split is exactly what the Open Responses project aims to fix. By providing a shared format, the goal is to prevent the market from becoming fragmented into different, incompatible silos. If more companies adopt a single standard, it becomes much easier for the entire AI industry to scale and for new tools to be integrated quickly.

The specification already enjoys significant support from a wide range of industry leaders. These include AWS (via Bedrock), Databricks, LM Studio, Ollama, Hugging Face, Nvidia, OpenRouter, Red Hat, and Vercel. Even open-source projects like vLLM and OGC have integrated the spec. While some major rivals like Anthropic and Google still have their own default APIs, their membership in the AAIF suggests a potential move toward these shared standards in the future.

What comes next for developers

For developers, the immediate benefit is the continued growth of the Open Responses ecosystem. Because the spec is now under the AAIF, it will benefit from a more diverse range of contributors and a more transparent development roadmap. This makes it a safer bet for companies looking to build long-term infrastructure that won't break if they decide to swap out their underlying AI model.

OpenAI has already contributed other pieces to this ecosystem, such as AGENTS.md, which provides a way to give coding agents specific instructions for a project. By donating Open Responses, OpenAI is positioning itself as a contributor to the foundational plumbing of the AI agent era, rather than just a provider of a single product. This shift toward open standards helps ensure that the tools of the future are interoperable and accessible to everyone.

The foundation will continue to provide a space where the technical specifications can be refined. As more providers join the standard, the 'translation' work that currently plagues AI development will slowly disappear, allowing for a more fluid movement of data and instructions across the entire AI landscape.