Major Technology companies collaborate with the Linux Foundation to establish standards for AI agents

Major technology companies have spent the last year insisting we are in the age of AI agents, yet much of what has been promised remains speculative. As organizations strive to bring imagination into reality, they have created a suite of tools to facilitate the evolution of generative AI. A group of significant participants in the AI competition, including Anthropic, Block, and OpenAI, has united to advocate for interoperability with the newly established Agentic AI Foundation (AAIF). This initiative promotes a select few established technologies and could establish them as a default benchmark for AI development in the future.

The trajectory for agentic AI models is uncertain at best, but companies have made such substantial investments in building these systems that some tools have emerged. The AAIF, affiliated with the nonprofit Linux Foundation, has been created to oversee the advancement of three pivotal AI technologies: Model Context Protocol (MCP), goose, and AGENTS.md.

MCP is likely the most recognized among the trio, having been made open-source by Anthropic a year prior. Its purpose is to connect AI agents to data sources in a consistent manner—Anthropic (and the AAIF now) refers to MCP as a “USB-C port for AI.” Instead of designing unique integrations for each different database or cloud platform, MCP enables developers to swiftly and effortlessly link to any MCP-compliant server.

Since its introduction, MCP has been extensively utilized throughout the AI sector. Google announced at I/O 2025 its intention to incorporate MCP support into its development tools, and numerous products have subsequently integrated MCP servers to enhance data accessibility for agents. OpenAI also embraced MCP just a few months following its launch.

The growing adoption of MCP may assist users in tailoring their AI interactions. For example, the new Pebble Index 01 ring features a local LLM capable of processing your voice notes, and it supports MCP for personalization.

Local AI models must compromise in comparison to larger cloud-based models, but MCP can bridge the functional gaps. “Many productivity and content tasks can be fully executed on the edge,” says Qualcomm’s head of AI products, Vinesh Sukumar, to Ars. “With MCP, you establish a connection with multiple cloud service providers for any complex task to be accomplished.”