Linux Foundation unveils Agentic AI standards group

The Linux Foundation has launched a new consortium, the Agentic AI Foundation (AAIF), with flagship contributions from Anthropic, Block and OpenAI in an effort to standardise how AI agents connect to tools, data and software projects.

The AAIF will steward three initial open source projects that target different layers of the emerging "agentic AI" stack: Anthropic's Model Context Protocol (MCP), Block's goose framework and OpenAI's AGENTS.md convention.

Jim Zemlin, Executive Director of the Linux Foundation, said the move comes as AI systems shift from simple conversational models towards agents that act on behalf of users and organisations.

"We are seeing AI enter a new phase, as conversational systems shift to autonomous agents that can work together. Within just one year, MCP, AGENTS.md and goose have become essential tools for developers building this new class of agentic technologies," said Zemlin. "Bringing these projects together under the AAIF ensures they can grow with the transparency and stability that only open governance provides."

Focus on shared standards

Agentic AI describes systems that can plan, make decisions and trigger actions across software and infrastructure with limited human intervention. Many of these systems rely on protocols and conventions that let models call tools, query data or navigate complex codebases.

The AAIF's structure mirrors other Linux Foundation-backed collaborations that sit at the infrastructure layer. It brings in large cloud providers, software vendors and AI specialists as members and maintainers of the core projects.

Platinum members of the new foundation include Amazon Web Services, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAI. Gold members include companies such as Cisco, Datadog, IBM, Oracle, Salesforce, SAP, Shopify and Snowflake. Silver members include firms such as Hugging Face, SUSE, Uber and Zapier.

Model Context Protocol

MCP, created by Anthropic, defines a standard way for AI models to connect with external tools, data sources and applications via servers that expose those resources. The protocol has become widely used over the past year across both proprietary and open-source AI products.

The Linux Foundation said more than 10,000 MCP servers have been published. These cover domains ranging from developer tooling to large corporate deployments. Major AI platforms, including Claude, Cursor, Microsoft Copilot, Gemini, VS Code and ChatGPT, have adopted the protocol.

Anthropic originally built MCP for internal use before releasing it under an open source license.

goose framework

Block has contributed goose, an open source "local-first" agent framework that it released earlier this year. The project combines language models with configurable tools and uses MCP-based integration.

Goose offers a structured environment for building and running AI workflows on local machines or within organisations. It focuses on predictable behaviour and controlled access to tools and data.

AGENTS.md convention

OpenAI has donated AGENTS.md, a markdown-based convention that gives AI coding agents a consistent way to read project-specific instructions across different code repositories and build systems.

The format sets out a standard place and structure for agent-oriented guidance within software projects. It is already in use across more than 60,000 open source projects and tools, including Amp, Codex, Cursor, Devin, Factory, Gemini CLI, GitHub Copilot, Jules and VS Code.

OpenAI has also contributed other components to the wider agentic ecosystem, such as the Codex CLI, Agents SDK and Apps SDK, which support MCP-based integrations.

Events and next steps

As part of the launch, AAIF member Obot.ai has donated its MCP Dev Summit events and related podcast to the new foundation. The next MCP Dev Summit is scheduled for New York City, with a European edition planned for later in the year.

The AAIF says it plans to expand its project portfolio and membership as the agentic AI market develops and as more vendors standardise on shared protocols for tools, data access and agent behaviour.