# State of MCP - Multi-agent, identities, and overlapping protocols > Source: > Published: 2026-09-18 17:37:07+00:00 Since its beginnings as a far-from-production-ready project at Anthropic connecting AI assistants to third-party data sources in late 2024, Model Context Protocol (MCP) has quickly grown to be a widespread open standard for connecting AI agents to enterprise systems. At the first [__AGNTCon/MCPCon Europe__](https://www.thestack.technology/aaif-sandbox-early-stage-projects/) since the protocol was [__donated as a founding project__](https://www.thestack.technology/new-foundation-openai-anthropic-block-ai-agent/) of the Linux Foundation’s Agentic AI Foundation (AAIF) in December, on Friday its creators, maintainers, and users looked back at that growth and ahead to the future of the protocol. “I think MCP today is in a pretty reasonable spot,” co-creator and Anthropic Technical Staff Member David Soria Parra told a keynote audience in Amsterdam, citing 500 million monthly downloads of the MCP tier-1 [__software-development kits__](https://modelcontextprotocol.io/docs/2026-07-28/sdk?ref=thestack.technology) (SDKs) and more than 1 billion MCP tool calls recorded through the Claude platform alone. Joining the AAIF has also helped it move beyond its roots as an Anthropic project; less than a quarter of its more than 40 maintainers work at the AI lab, and its core team includes maintainers from Google Cloud, Microsoft, OpenAI, and AWS. “You cannot deploy agentic applications at scale without MCP, that needs to become a central part of every deployment,” AAIF Executive Director Mazin Gilbert told *The Stack* during a conversation at the event.