Agentic AI standards race to catch up with rapid industry adoption

Mazin Gilbert, executive director of the Linux Foundation’s Agentic AI Foundation (AAIF), is driving a rapid standardization effort for agentic AI. Speaking at AGNTCon + MCPCon in San Jose last month, he emphasized that the foundation must operate at extreme speed to remain relevant as major tech companies build products both with and without emerging standards.

What happened

Gilbert left Google earlier this year to lead the AAIF, which has since become the fastest-growing foundation in the Linux Foundation’s history. The organization now counts 305 members, with approximately one new company joining every business day. About 40 percent of these members are startups, while 19 percent are nonprofits, national labs, or academic institutions. The recent conference in San Jose attracted 4,300 registrants, doubling the attendance of the previous largest event in Amsterdam.

The foundation faces a diverse membership landscape. Some contributors actively ship products based on shared protocols, while others are still attempting to move agents from experimental labs into production environments. Gilbert noted that a third group of companies is effectively panicking because they lack a clear agent strategy. To address this, the AAIF launched its first certification, the Model Context Protocol Associate (MCPA), in September to help developers gain necessary skills. The Gates Foundation joined the board in September to support training and certification access for developers in Africa and South Asia.

Collaboration among fierce competitors remains central to the mission. Gilbert described meetings where maintainers from rival companies worked together on protocols because they all depend on them for infrastructure. He stated that these protocols are like plumbing: unglamorous but essential for preventing a messy, expensive, and prohibitive deployment landscape for enterprises.

How it works

The core technical driver for this standardization is the Model Context Protocol (MCP). Anthropic donated the protocol to the foundation last December, and in July, the specification was updated to be stateless. This change required contributing companies to update their entire stacks, demonstrating a high level of commitment. Since its donation, MCP downloads have grown tenfold in the first ten months. Today, fewer than 15 percent of its maintainers work at Anthropic, with over 1,000 organizations contributing 75 percent of the pull requests.

The AAIF acts as a neutral governance body to evaluate and host projects. It uses a public intake process to review submissions, having declined some to protect the foundation's direction. The technical committee and governing board assess proposals based on whether the AAIF is the right home for the technology and how it fits into layers such as agent frameworks, gateways, guardrails, memory, and observability. The goal is to harmonize competing efforts, such as the multiple gateway projects currently under consideration, rather than fragmenting the ecosystem further.

Key details

• The AAIF has 305 members, growing by roughly one company per business day.
• MCP downloads increased tenfold in the first ten months after Anthropic donated the protocol.
• Over 1,000 organizations now contribute 75 percent of MCP pull requests, with less than 15 percent of maintainers from Anthropic.
• The foundation currently hosts 10 projects, including Agent Skills, Open Responses, OpenEnv, and OpenSandbox.
• Two gateway projects, Solo.io’s agentgateway and Agent Router, were approved this year, with several more pending harmonization decisions.
• The Model Context Protocol Associate (MCPA) certification launched in September to aid developer training.

Why it matters

For software engineers and technical leaders, the shift toward standardized protocols like MCP reduces the burden of maintaining custom integrations. Without common standards, companies must write and support proprietary code to connect agents to backend systems, leading to higher costs and fragility. As Gilbert noted, if the plumbing is not correct, the entire agentic AI space becomes prohibitively expensive for scale. The move toward stateless protocols and shared gateways allows teams to focus on building unique product value rather than reinventing basic connectivity.

The rapid growth of the AAIF signals that agentic AI is moving from experimentation to enterprise infrastructure. Engineers need to understand that participating in these working groups is no longer optional for staying current. Whether through contributing code or adopting certifications like MCPA, engaging with these standards helps ensure that future agent deployments are interoperable, secure, and transactional. The foundation’s focus on identity, security, and eventually agent commerce suggests that the next wave of development will require robust, standardized trust layers.

What you can do

• Join an AAIF working group to listen and learn, even if you are not yet an expert in agent architecture.
• Evaluate your current agent integrations for compatibility with the stateless Model Context Protocol.
• Consider pursuing the Model Context Protocol Associate (MCPA) certification to validate your team’s skills.
• Monitor the foundation’s project intake decisions to identify which gateway or framework solutions are gaining community consensus.
• Assess your organization’s agent strategy to determine if you are in the building, learning, or planning phase.
• Contribute to open-source MCP repositories if your team relies on these protocols for production systems.