Today marks a major milestone in the evolution of agentic AI. OpenAI has joined hands with Anthropic and Block to co-found the Agentic AI Foundation (AAIF) under the Linux Foundation. With support from major tech leaders—including Google, Microsoft, AWS, Bloomberg, and Cloudflare—the initiative aims to build neutral, open, and interoperable infrastructure for next-generation AI agents.
As AI moves from experimentation to real-world use, the industry needs shared standards more than ever. The AAIF steps in at this crucial moment to guide the transition.
Why Open Standards Matter for Emerging Agentic Systems
Developers around the world are rapidly adopting AI to create more capable agentic systems. These agents now power coding workflows, automation pipelines, customer support, and daily business operations. Because these tools now handle real responsibility, consistent standards are essential.
Open standards help in several ways:
They improve interoperability across tools and platforms.
They enhance safety, since agents follow consistent conventions.
They boost portability, allowing developers to move agents across environments easily.
They prevent fragmentation as this new category of AI matures.
Without alignment, agent development could splinter into isolated ecosystems. That fragmentation would slow progress, reduce safety, and limit innovation. Therefore, AAIF was created to guide the ecosystem toward collaboration.
Building an Open Agentic Ecosystem
Over the past year, OpenAI has invested heavily in open agentic infrastructure. These contributions include:
Agents SDK
Apps SDK
Agentic Commerce Protocol
gpt-oss models
Codex CLI
Codex alone has helped merge over two million public pull requests on GitHub, showing the real-world power of shared, open tools.
OpenAI has also been a leading contributor to the Model Context Protocol (MCP). This protocol now forms the backbone for connectors and apps inside ChatGPT. Recently, OpenAI, Anthropic, and MCP-UI partnered to extend the Apps SDK to the broader developer community through MCP Apps, making it easier to build cross-platform, agent-powered experiences.
These efforts laid the foundation for AAIF by proving that open, interoperable agent infrastructure can succeed at scale.
Donating AGENTS.md: A Step Toward Universal Agent Instructions
OpenAI has also contributed AGENTS.md to the AAIF. This lightweight Markdown file was developed to solve a common problem: how can agents reliably find project-specific instructions such as coding conventions, build steps, or testing rules?
AGENTS.md lives alongside a project’s README.md and gives agents predictable, structured guidance. Since its launch in August 2025, more than 60,000 open-source projects have adopted it. Popular tools and frameworks—including Amp, Codex, Cursor, Devin, Factory, Gemini CLI, GitHub Copilot, Jules, and VS Code—now rely on its conventions.
By donating AGENTS.md to the AAIF, OpenAI ensures that:
Developers get a portable, community-driven standard
The format evolves openly
No single company controls its future
Additionally, AAIF co-founders contribute other key projects such as Anthropic’s Model Context Protocol (MCP) and Block’s goose.
What the Agentic AI Foundation Enables
The AAIF provides a neutral home for developing and governing interoperability standards for agentic systems. Operating under the Linux Foundation—a trusted steward behind the Linux Kernel, Kubernetes, Node.js, and PyTorch—the foundation ensures long-term sustainability and transparent governance.
Established by OpenAI, Anthropic, Block, Google, Microsoft, Amazon, Bloomberg, and Cloudflare, the AAIF aims to accelerate agentic AI innovation through:
Shared investment
Open protocols
Community-led standards
Transparent decision-making
As agentic AI continues to evolve, the AAIF will help ensure it grows in a way that benefits developers, enterprises, and the broader public.