Agentic AI Foundation Has Over 250 Member Companies, Brinkmann Says

The Agentic AI Foundation has more than 250 member companies, has received over 1,000 ambassador applications, and has community chapters in more than 80 cities, Demetrios Brinkmann said in a newly published Practical AI interview. The figures describe the scale of participation in the foundation that now stewards the Model Context Protocol, following Anthropic’s previously announced donation of the standard.

MCP is an open standard for connecting AI models with outside data sources and tools. Anthropic had previously donated the protocol to the Linux Foundation’s Agentic AI Foundation, saying it should belong to the broader community. Brinkmann described that earlier change as a move from Anthropic ownership to stewardship by a neutral governing body.

That distinction is more than organizational housekeeping. A protocol determines how a model reaches information and software beyond its own chat window. Anthropic argues that open protocols can build security properties into shared infrastructure instead of having each deployment assemble its own approach. It also says the model providers should compete on the quality and safety of their agents, rather than on control of integrations.

More than 250Member companies

Brinkmann said the Agentic AI Foundation had more than 250 member companies at the time of the interview.

More than 1,000Ambassador applications

Brinkmann said more than 1,000 people applied to the foundation’s ambassador program.

More than 80 citiesCommunity chapters

Brinkmann said the community had chapters in more than 80 cities and about 250 chapter leads.

The figures Brinkmann cited show organizations and individuals seeking a role in the foundation’s community. They do not, on their own, establish how widely MCP is deployed, whether implementations work consistently, or whether users receive the security benefits Anthropic describes. Those questions sit beyond membership totals and event networks.

Foundation governance does not remove the risks created when agents reach into tools and data. Anthropic warns that agents with less human oversight can misunderstand intent or take unintended actions, and that prompt-injection attacks can try to manipulate an agent through the content it processes. Its position is that open infrastructure complements, rather than replaces, the security work required from model developers and customers.