MCP adoption surges as vendors ship connectors and developer tooling, dubbed the 'USB‑C for AI agents'

Model Context Protocol is moving from an Anthropic project to shared infrastructure for AI agents, with Anthropic saying the standard now spans more than 10,000 public servers. (anthropic.com) Anthropic donated MCP to the Linux Foundation’s Agentic AI Foundation on December 9, 2025, alongside Block and OpenAI as co-founders, with support from Google, Microsoft, Amazon Web Services, Cloudflare and Bloomberg. (anthropic.com) In that announcement, Anthropic said MCP had already been adopted by ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code, and that official SDK downloads across Python and TypeScript had reached 97 million per month. (anthropic.com) MCP is a shared language for connecting a model to outside tools and data, the way USB-C gives devices one port instead of a different cable for every accessory. The official documentation uses that USB-C analogy directly. (modelcontextprotocol.io) The pitch is simple: build one client or server once, then reuse it across many agent products instead of writing a custom integration for every model-tool pairing. Anthropic’s engineering docs describe MCP as a universal protocol for that problem. (anthropic.com) OpenAI is now documenting MCP as well, calling it an open protocol that is becoming the industry standard for extending models with tools and knowledge. Its developer docs show how to build remote MCP servers for ChatGPT apps, deep research and API integrations. (openai.com) That shift is showing up in vendor tooling. Composio has published step-by-step MCP integration guides for its own platform with the OpenAI Agents SDK, plus toolkit-specific guides for Botpress with Pydantic AI and Polygon with the OpenAI Agents SDK. (composio.dev 1) (composio.dev 2) (composio.dev 3) Those guides are concrete rather than theoretical. The Botpress example walks through listing conversations, bot issues and workspace files through natural-language commands, while the Polygon example covers market status, stock tickers and crypto indicators. (composio.dev 1) (composio.dev 2) The operational work is catching up with the adoption curve. The MCP roadmap says production deployments have exposed gaps around horizontal scaling, stateless operation, session migration behind load balancers, retry semantics and task expiry policies. (modelcontextprotocol.io) The maintainers are also expanding governance as the protocol grows. An April 8, 2026 post said the project had already gone through two specification releases, the move to the Agentic AI Foundation and a steady increase in proposal volume. (blog.modelcontextprotocol.io) The result is that MCP is no longer just a protocol spec on a documentation site. It is becoming the connector layer that major model vendors, cloud platforms and integration companies are all trying to standardize around. (anthropic.com) (openai.com)