The three most important open standards for AI agents, Anthropic’s Model Context Protocol, OpenAI’s AGENTS.md, and Block’s Goose, are now governed by the same neutral organization. The Agentic AI Foundation (AAIF) launched in December 2025 under the Linux Foundation with 49 member organizations, including every major cloud provider, every major AI lab, and enterprise vendors from SAP to Salesforce. If you build or deploy AI agents, the standards that define how your agents connect to tools, read project instructions, and execute workflows are no longer controlled by any single company.
That matters more than any new framework release this year.
Three Projects, Three Problems Solved
The AAIF is not a lobbying group or a marketing consortium. It governs three specific open-source projects, each solving a different piece of the agent infrastructure puzzle.
MCP: The Universal Tool Connector
The Model Context Protocol (MCP) standardizes how AI agents connect to external tools, data sources, and services. Anthropic created it in November 2024. By the time it was donated to the AAIF, MCP had reached 10,000+ published servers, 97 million monthly SDK downloads across Python and TypeScript, and adoption by ChatGPT, Cursor, Gemini, Microsoft Copilot, and VS Code.
Before MCP, connecting an agent to your GitHub repo, Postgres database, and Slack workspace meant writing three separate integrations with three different auth flows and data formats. MCP replaces that with a single protocol: one interface for any tool.
The donation to AAIF means Anthropic no longer unilaterally controls MCP’s roadmap. A steering committee with representatives from multiple member organizations now guides technical decisions. The immediate roadmap includes asynchronous operations, statelessness support, and official SDKs in all major programming languages.
AGENTS.md: The README for Coding Agents
AGENTS.md is a markdown file you put at the root of a repository. It tells AI coding agents what they need to know to work on your project: build commands, test procedures, coding conventions, security rules, directory structure quirks. Think of it as a README that is written for machines instead of humans.
OpenAI released AGENTS.md in August 2025. By December, 60,000+ open-source projects had adopted it. Coding agents from Cursor, Codex, Devin, Factory, Gemini CLI, GitHub Copilot, Jules, and VS Code all read it. The format is intentionally simple. No schema validation, no YAML headers, just markdown with structured sections that agents parse from the nearest file in the directory tree.
The practical impact: a coding agent working on a monorepo can read different AGENTS.md files for different subprojects. Each subproject ships its own instructions. The agent adapts without prompt engineering from the user.
Goose: The Local-First Agent Framework
Goose is Block’s open-source agent framework. It is the least known of the three but arguably the most opinionated. Goose runs locally, works with any LLM, and treats MCP as its primary integration layer. It can build projects from scratch, write and execute code, debug failures, and orchestrate multi-step workflows.
Block released Goose under Apache 2.0 in early 2025. It serves as the reference implementation for MCP, which means if you want to see how MCP is supposed to work in a production agent, Goose is the example. Its donation to AAIF ensures that the reference implementation stays vendor-neutral alongside the protocol itself.
Who Controls the Standards Now
The AAIF operates as a directed fund under the Linux Foundation, the same organizational model that governs Kubernetes, Node.js, and PyTorch. Members pay into a common fund. That money finances engineering, infrastructure, marketing, and community programs. Technical decisions stay with project maintainers and steering committees.
The membership roster reads like an industry census:
Platinum (8): AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, OpenAI
Gold (19): Adyen, Arcade.dev, Cisco, Datadog, Docker, Ericsson, IBM, JetBrains, Okta, Oracle, Runlayer, Salesforce, SAP, Shopify, Snowflake, Temporal, Tetrate, Twilio
Silver (23): Apify, Chronosphere, Cosmonic, Elasticsearch, Eve Security, Hugging Face, Kubermatic, KYXStart, LanceDB, Mirantis, NinjaTech AI, Obot.ai, Prefect.io, Pydantic, Shinkai.com, Solo.io, Spectro Cloud, Stacklok, SUSE, Uber, WorkOS, Zapier, ZED
That is not just AI companies. Payment processors (Adyen, Block), DevOps vendors (Datadog, Docker, JetBrains), enterprise software giants (SAP, Salesforce, Oracle), and telecom companies (Ericsson, Twilio) are all in the room. When standards organizations attract participants from outside their core domain, the standards tend to stick.
The Platinum Paywall Concern
Not everyone is thrilled. Critics have pointed out that the AAIF’s governance model gives platinum members disproportionate influence over funding allocation and strategic direction. The Linux Foundation model is well-tested, but “open governance” means something different when a platinum seat costs significantly more than silver membership. Individual contributors and smaller organizations may find their voices diluted.
That said, the Linux Foundation’s track record with Kubernetes and other projects shows that technical merit tends to win over politics in steering committees. The projects themselves remain under open-source licenses (Apache 2.0), so anyone can fork if governance fails.
What This Changes for Builders
If you are building AI agents today, three things change immediately.
MCP is no longer “Anthropic’s protocol.” Enterprise procurement teams that hesitated to adopt MCP because of single-vendor concerns now have a neutral governance story. AWS, Google Cloud, and Microsoft Azure are all building MCP infrastructure. The protocol is on a path to becoming the HTTP of agent-tool communication.
AGENTS.md becomes a project hygiene standard. With 60,000 projects already using it and every major coding agent supporting it, skipping AGENTS.md in your repository is like skipping a README in 2015. It costs you nothing to add and makes every agent that touches your code more effective. If you maintain open-source projects, add one today.
Framework choices become less permanent. Goose, as the MCP reference implementation under neutral governance, sets a baseline that other frameworks will follow. If your agent framework supports MCP (and most do, since LangChain, CrewAI, and Pydantic AI all support it), you can swap frameworks without rewriting your tool integrations.
The first MCP Dev Summit is scheduled for April 2-3, 2026 in New York City. Expect the roadmap for MCP 2.0 features, including asynchronous operations and streamlined authentication, to become clearer there.
What This Changes for Enterprises
For enterprise buyers, AAIF simplifies the “which standard do we bet on?” question. You bet on all three because they are now one foundation.
Compliance gets easier. When your auditor asks who controls the protocol your agents use to access production databases, “the Linux Foundation, same organization that governs Kubernetes” is a stronger answer than “Anthropic, but they promise to keep it open.” Neutral governance removes a procurement risk.
Vendor lock-in decreases. If your agent infrastructure runs on MCP, and MCP is governed by a foundation with AWS, Google, and Microsoft as equal platinum members, no single vendor can change the protocol to disadvantage competitors. That is the same dynamic that made Kubernetes the standard for container orchestration.
Interoperability becomes the default. With Salesforce, SAP, Oracle, and Shopify as AAIF members, expect MCP connectors for major enterprise platforms to proliferate. The pattern is familiar: open standards attract integrations because vendors can invest in connectors without worrying about protocol stability.
The risk? Standards bodies can also slow things down. MCP moved fast under Anthropic’s sole stewardship. Committee-driven development can introduce politics and delay features. The Linux Foundation’s experience helps here, but it is worth watching how quickly the post-AAIF roadmap progresses compared to MCP’s first year.
Frequently Asked Questions
What is the Agentic AI Foundation (AAIF)?
The Agentic AI Foundation is a directed fund under the Linux Foundation that governs three open-source AI agent projects: Anthropic’s Model Context Protocol (MCP), OpenAI’s AGENTS.md, and Block’s Goose. It was founded in December 2025 with 49 member organizations including AWS, Google, Microsoft, and OpenAI.
Who are the founding members of the Agentic AI Foundation?
The three founding contributors are Anthropic (which donated MCP), OpenAI (which donated AGENTS.md), and Block (which donated Goose). Platinum members also include AWS, Bloomberg, Cloudflare, Google, and Microsoft.
What does the Agentic AI Foundation mean for MCP?
MCP is now governed by a neutral steering committee instead of Anthropic alone. The protocol keeps its open-source license (Apache 2.0), but roadmap decisions are now made by representatives from multiple member organizations. The upcoming roadmap includes asynchronous operations, statelessness support, and SDKs in all major programming languages.
What is AGENTS.md and why does it matter?
AGENTS.md is a markdown file placed in the root of a code repository that tells AI coding agents how to work on the project. It includes build commands, test procedures, and coding conventions. Over 60,000 open-source projects have adopted it, and all major coding agents (Cursor, Codex, GitHub Copilot, Devin) support it.
How does the AAIF affect enterprise AI agent deployments?
The AAIF reduces vendor lock-in risk because MCP, the primary protocol for agent-tool connections, is now governed by a neutral foundation rather than a single company. Enterprise procurement teams can adopt MCP with confidence that no single vendor controls the standard, similar to how Kubernetes is governed for container orchestration.