Today marks a significant milestone: the Model Context Protocol (MCP) joined the Linux Foundation’s new Agentic AI Foundation, with support from Microsoft and others. Alongside MCP, Anthropic is joined by Block (donating Goose, their agent framework) and OpenAI (contributing AGENTS.md).
This isn’t just a code donation. It’s MCP graduating from a promising protocol into foundational infrastructure for the agent era.
As part of the MCPB working group, I’ve watched this ecosystem evolve over the past months. Today’s announcement represents the culmination of careful planning to ensure MCP can scale as neutral, vendor-independent infrastructure.
The Problem Every Open Protocol Faces
When Anthropic built MCP, they created something technically elegant: a standard way to connect AI models and agents to tools and data. The protocol worked. Adoption grew. Developers started building on it.
Then came the governance questions:
- Who owns the roadmap? When multiple companies have competing priorities, who decides what gets built next?
- How do we balance velocity with consensus? Move too fast, and you alienate contributors. Move too slow, and the ecosystem fragments.
- When does one company’s urgency become everyone’s problem? Every stakeholder thinks their use case is the most important.
These aren’t technical problems. They’re ecosystem coordination challenges. And they require completely different skills to solve.
Den Delimarsky helped push MCP to have well-defined governance from the get-go, recognizing early that sustainable protocols need clear decision-making structures.
Why Governance Matters
Protocols don’t fail because the tech is wrong. They fail because the governance doesn’t scale.
The pattern repeats: a company builds a promising protocol, others adopt it, adoption creates pressure to evolve, but governance fails to accommodate diverse stakeholders. The ecosystem fragments or dies.
MCP needed to break this pattern.
What the Linux Foundation Changes
MCP joining the Agentic AI Foundation isn’t about moving code—it’s about establishing governance and funding that can scale with neutral oversight.
Here’s what changes:
Neutral Infrastructure
Before: Anthropic owned MCP. No matter how open the code, the perception was “Anthropic’s protocol.”
After: The Linux Foundation provides neutral infrastructure. No single vendor controls the roadmap. Technical steering committees make decisions based on merit and community consensus.
Shared Standards, Shared Benefits
Jim Zemlin, executive director of the Linux Foundation, put it bluntly: the goal is to avoid a future of “closed wall” proprietary stacks.
When MCP becomes neutral infrastructure, contributions benefit the entire ecosystem. Improvements from any contributor flow back to everyone building agent systems.
This changes the incentive structure. Companies can invest in shared standards with confidence that the governance will remain neutral.
Evolution Without Fragmentation
MCP co-creator David Soria Parra told TechCrunch: “The main goal is to have enough adoption in the world that it’s the de facto standard. We’re all better off if we have an open integration center where you can build something once as a developer and use it across any client.”
That vision requires governance that supports evolution without control. The Linux Foundation’s model—technical steering committees, community-driven roadmaps, transparent decision-making—enables that.
What This Means for the Agent Ecosystem
The Agentic AI Foundation isn’t just about MCP. It’s anchoring three critical pieces of agent infrastructure:
- MCP (Anthropic) - Standard protocol for connecting models/agents to tools and data
- Goose (Block) - Open-source agent framework
- AGENTS.md (OpenAI) - Simple instruction files for AI coding tools
Together, these represent the basic plumbing of the agent era. When they’re governed neutrally, the entire ecosystem can build on shared standards.
Other members include AWS, Bloomberg, Cloudflare, Google, and Microsoft. This isn’t a vendor alliance—it’s an industry acknowledgment that shared infrastructure requires shared governance.
What Comes Next
This is just the beginning. Success will be measured by:
- Adoption across competing platforms - MCP becoming the de facto standard for agent-tool connections
- Evolution velocity - Technical steering committees keeping pace with rapid AI development
- Ecosystem growth - Developers building with confidence on stable, neutral standards
OpenAI engineer Nick Cooper captured the vision: “I don’t want it to be a stagnant thing. I don’t want these protocols to be part of this foundation, and that’s where they sat for two years. They should evolve and continually accept further input.”
The opportunity: governance that provides stability without stagnation, consensus without gridlock, openness without chaos.
Today’s announcement positions MCP to become exactly what the agent ecosystem needs: neutral infrastructure that lets everyone build higher.
The Lesson
If you’re building a protocol that could become infrastructure: plan for governance from day one. Don’t wait until adoption creates governance pressure—by then, stakeholders have incompatible expectations.
The technical work is table stakes. The governance work determines whether your protocol becomes infrastructure or becomes irrelevant.
I contribute to the MCPB (MCP Bundle) working group and have been working with the MCP ecosystem for several months. These reflections are my own observations about open protocol governance.