The heavyweights of artificial intelligence are taking the lead to shape the nascent market for AI agents.
OpenAI, Anthropic, and Block announced the creation of the Agentic AI Foundation (AAIF), a new foundation hosted by the Linux Foundation. The aim: to establish open standards for these autonomously acting systems, before market fragmentation takes root.
Among its members are Cloudflare, Oracle, Cisco, IBM, Salesforce.
The initiative doesn’t emerge from nowhere. Building on the Linux Foundation, the three founders are adopting a governance model proven in open source: decision transparency, public rules, and balanced representation of contributors. A choice intended to reassure companies wary of depending on a single vendor for technologies so critical.
The stakes are high. After years of experimentation, AI agents are leaving the lab to become a standalone enterprise infrastructure. These systems, capable of planning and executing complex tasks with minimal human supervision, raise major questions about security, interoperability, and vendor lock-in.
Three Core Technological Pillars at the Heart of the Framework
The AAIF begins with tangible building blocks, anchored by three open-source projects already widely adopted by developers. Anthropic contributes the Model Context Protocol (MCP), a protocol that standardizes the connection of language models to tools and external systems. A fundamental component for orchestrating agents capable of acting in complex environments.
Meanwhile, Block contributes goose, a modular framework designed to build and deploy extensible AI agents.
OpenAI, finally, brings AGENTS.md into the foundation’s fold. This open instruction format, already used by tens of thousands of projects, works as a “machine README”: it documents the capabilities, tools, and behaviors of agents to facilitate interoperability.
Backing from the Cloud Giants
The initiative immediately benefits from heavyweight support. AWS, Google, Microsoft, Bloomberg, and Cloudflare have joined the foundation as platinum members. This setup gives the AAIF immediate influence over shaping de facto standards for agentic AI, while signaling neutrality toward providers.
For developers, the promise is a unified baseline of protocols and formats to build agents that operate across different clouds, tools, and code repositories. By standardizing technologies already widely used, the foundation aims to reduce integration costs and accelerate enterprise adoption.
Beyond interoperability, the AAIF emphasizes safety and reliability. By establishing a common framework to describe agents’ capabilities, control their access to tools, and coordinate their actions across systems, the foundation seeks to mitigate risks tied to ad hoc implementations. The objective: facilitate auditing of agent behaviors and collectively evolve best practices as these technologies diffuse.
