David Soria Parra and Justin Spahr-Summers published a spec at Anthropic on Nov. 25, 20241. It solved a problem every AI lab shared and every lab had left alone. How does an assistant reach the data and tools sitting outside its own weights?
Thirteen months later, on Dec. 9, 2025, Anthropic handed that spec to a brand-new nonprofit2.
What each protocol actually standardizes
The two solve nearby problems, and mixing them up hides why both had to exist. MCP governs the link between an agent and its tools1. Databases, file systems, calendars, codebases: anything an assistant reads or writes outside its own context. Anthropic's framing set the goal as replacing one-off integrations with a single protocol a developer writes once and reuses. A2A governs the link between agents instead6. Agents built on different frameworks find each other's capabilities, agree how to interact, and work a task together while each keeps its own state private.
Picture a travel-booking agent that needs a hotel-search agent, a currency agent and a calendar agent to finish one itinerary. MCP connects each agent to its own data. A2A connects the agents to each other. Running agents at scale needs both layers at once. That is why overlapping groups govern them, rather than rival standards bodies.
A foundation built to hold three donations
Jim Zemlin, the Linux Foundation's executive director, framed the purpose around what changed. "We are seeing AI enter a new phase, as conversational systems shift to autonomous agents that can work together."2 He tied the governance to that shift. "Bringing these projects together under the AAIF ensures they can grow with the transparency and stability that only open governance provides."
Mike Krieger, Anthropic's chief product officer, represented MCP's transfer and described where it came from. "MCP started as an internal project to solve a problem our own teams were facing."2 He named the reason for giving it away. "Donating MCP to the Linux Foundation as part of the AAIF ensures it stays open, neutral, and community-driven as it becomes critical infrastructure for AI." Manik Surtani, Block's head of open source, put the stakes wider. "We're at a critical moment for AI. The technology that will define the next decade, that promises to be the biggest engine of economic growth since the Internet, can either remain closed and proprietary for the benefit of few, or be driven by open standards, open protocols, and open access for the benefit of all." Nick Cooper, a member of OpenAI's technical staff, spoke for AGENTS.md in plainer terms. "For AI agents to reach their full potential, developers and enterprises need trustworthy infrastructure and accessible tools to build on."
Eight platinum members anchored the launch: Amazon Web Services, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft and OpenAI28. Each sent a named technical executive. Swami Sivasubramanian of AWS joined Shawn Edwards of Bloomberg, Dane Knecht of Cloudflare, Richard Seroter of Google Cloud and Chris DiBona of Microsoft. Gold members filled out the list with Cisco, Docker, IBM, Oracle, Salesforce, SAP, Shopify and Twilio. Hugging Face, Uber, Zapier and eighteen more joined at silver.
Numbers in the announcement measured real use rather than hope. More than 10,000 MCP servers were already published when Anthropic handed the protocol over2. Some 60,000 open-source projects had adopted AGENTS.md as their agent-guidance format. Growth sped up after that345. The foundation added 97 members in February 2026, 43 in May, and 57 in August, pulling in big financial firms and Asia-Pacific tech leaders. Three waves in eight months carried the roster well past the roughly 40 companies at launch. Read it as buyers deciding the standard was safe to build on.
The Agentic AI Foundation added 197 members in three 2026 waves, led by 97 in February
Companies joining the Agentic AI Foundation at its December 2025 launch and in each 2026 wave, count of members
Each column counts the companies that joined at one step: the blue column is the roughly 40 founding companies, and the magenta columns are the three waves announced in 2026.
Sources: Linux Foundation [2]; Linux Foundation [3]; Linux Foundation [4]; Linux Foundation [5]. Chart by Kelly Dennis.
The numbers behind this chart
| Item | Value | Note |
|---|---|---|
| Launch, Dec. 2025 | 40 | Roughly 40 companies, anchored by eight platinum members |
| Feb. 2026 | 97 | |
| May 2026 | 43 | |
| Aug. 2026 | 57 | Big financial firms and Asia-Pacific tech leaders |
Google keeps A2A on a parallel track
A2A's governance runs beside the foundation's, on purpose. Google gave the protocol to the Linux Foundation as its own project under an Apache license6. That kept it apart from the three gifts that formed the Agentic AI Foundation, though the two efforts share plenty of members. IBM Research co-built it with Google Cloud, which gives A2A an enterprise-research origin, where MCP came out of developer tooling. By this research's count, the repository had 25,600 stars and 2,600 forks, with kits in Python, Go, JavaScript, Java, .NET and Rust. That reach goes well past one language community. Staying separate let Google keep a visible credit for the standard, while giving up the control a single-vendor spec carries.
What standardization implies for platform power
Handing a protocol to a neutral body looks like giving up leverage. Look again. It keeps a different kind. Whichever company's engineers sit closest to a technical committee shape the interfaces every rival has to support. Krieger's role for MCP inside the foundation, beside Seroter's for Google Cloud and DiBona's for Microsoft, puts the three firms racing hardest on agent products closest to the rules2.
Masaaki Hirano, founder and chief executive of Wintermute Technologies, posted his read of the same donation to the developer platform Dev.to on Dec. 9, 2025. "They will standardize the troublesome part of 'connection,' but will then shift all their resources to locking you in with 'Intelligence' and 'Ecosystems' beyond that connection," he wrote7. A shared protocol for data access still leaves each platform free to sell its own proprietary reasoning layer on top of it.
Buyers gain either way. An open protocol lowers the cost of switching vendors, and cuts the odds one platform traps a customer through its own tool-calling format. Cost adds urgency. Gartner's Aug. 10, 2026 forecast projected 96% growth in AI-ready cloud spending this year. That number works only if agents on one stack can call tools and agents on a rival's, while the integration work stays small. Shared standards are what make the spending productive rather than just bigger.
Security worries follow every standard that moves this fast, and this one carries a sharper version. A protocol that decides which tools an agent can call, and which agents it trusts, becomes a target the moment enterprises adopt it. Prompt injection and agent-hijacking research already treats MCP-style tool access as a main route in, a concern this edition covers separately. Neutral governance helps here too. A foundation whose top members compete with each other has more reason to patch flaws in the open than a lone vendor facing a disclosure that hurts its own product.
By the numbers
- Nov. 25, 2024: the day Anthropic published the Model Context Protocol specification1.
- Dec. 9, 2025: the day the Linux Foundation formed the Agentic AI Foundation around MCP, goose and AGENTS.md2.
- Eight platinum members anchored the foundation at launch, among them AWS, Google, Microsoft and OpenAI2.
- More than 10,000 MCP servers had been published by the foundation's founding date2.
- Sixty thousand-plus open-source projects had adopted AGENTS.md as their agent-guidance format2.
- A combined 197 new members joined across three 2026 waves, 97 arriving in February, then 43, then 57345.
- GitHub stars on Google's Agent2Agent repository stood at 25,600 under Linux Foundation stewardship6.
- Six programming languages have published A2A software-development kits6.
What to watch
Membership growth measures interest, so the sharper signal comes from production. Watch for enterprises publishing case studies of MCP servers and A2A agents running live work rather than pilots. Do the two protocols formally reference each other in updated specs? Agents that call tools and coordinate with peers need both layers to fit cleanly. Security audits and a published disclosure record from the Agentic AI Foundation would give buyers the evidence to move past pilot budgets.
Sources
- David Soria Parra and Justin Spahr-Summers, "Introducing the Model Context Protocol," Anthropic, Nov. 25, 2024, https://www.anthropic.com/news/model-context-protocol.
- "Linux Foundation Announces the Formation of the Agentic AI Foundation (AAIF), Anchored by New Project Contributions Including Model Context Protocol (MCP), goose and AGENTS.md," Linux Foundation, Dec. 9, 2025, https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation.
- "Agentic AI Foundation Welcomes 97 New Members As Demand for Open, Collaborative Agent Standardization Increases," Linux Foundation, Feb. 24, 2026, https://www.linuxfoundation.org/press.
- "Agentic AI Foundation Adds 43 New Members as Enterprise and Government Adoption of Open Agent Standards Accelerates," Linux Foundation, May 18, 2026, https://www.linuxfoundation.org/press.
- "Agentic AI Foundation Welcomes 57 New Members, Gaining Major Financial Services Players and APAC Leaders," Linux Foundation, Aug. 12, 2026, https://www.linuxfoundation.org/press.
- "a2aproject/A2A," GitHub, A2A Project, 2026, https://github.com/a2aproject/A2A.
- Masaaki Hirano, "MCP Wins, But the War Just Started: Why Standardization Won't Save Us from Lock-in," DEV Community, Dec. 9, 2025, https://dev.to/fl4tlin3/mcp-wins-but-the-war-just-started-why-standardization-wont-save-us-from-lock-in-53a4.
- Makenzie Holland, "Big Tech Takes Steps to Build Open Standards for Agentic AI," CIO Dive, Dec. 10, 2025, https://www.ciodive.com/news/big-tech-develop-open-standards-agentic-ai/807608/.
