LEETConsulting
← Insights

2 min readLeet Consulting

MCP goes to a foundation. Why that matters for enterprise integration

The Model Context Protocol now lives at the Linux Foundation, backed by Anthropic, OpenAI and Block. For enterprises, the question shifts from "which model?" to "which systems can agents safely reach?"

  • Platforms
  • Agents

On December 9, Anthropic donated the Model Context Protocol to the new Agentic AI Foundation, a directed fund under the Linux Foundation co-founded with Block and OpenAI. Anthropic reports more than 10,000 active public MCP servers and over 97 million monthly SDK downloads, with adoption in ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code.

For people who build integrations, this is the most important news of the quarter, more than any single model release.

Why a neutral standard matters

Every agent project we have scoped spends more time on integration than on prompts. The agent has to read the policy system, update the claim, look up the provider, file the note. Before MCP, each of those was a custom tool for one framework and one model vendor.

A protocol under neutral governance means a connector you build once can be used by agents on different models. The same quarter made that concrete:

  • In October, OpenAI launched AgentKit with a Connector Registry, an agent builder and evals for agents.
  • In November, Google released Gemini 3 alongside Antigravity, an agentic development platform working across editor, terminal and browser.

Three major platforms, each with its own agent tooling, converging on one way to connect to your systems.

How we advise clients to use it

  • Treat MCP servers as products. Each one is an API into a system of record. It needs an owner, versioning, authentication and logs.
  • Expose verbs, not databases. "Get claim status" and "add claim note" are safer and easier to evaluate than raw SQL access.
  • Put permissions in the server. Never rely on the prompt to stop an agent from doing something. The server should not offer what the agent may not do.
  • Keep models swappable. With a standard connector layer, changing model vendors becomes an evaluation exercise rather than a rebuild.

Model releases will keep coming every few weeks. The connectors you build now are what will still be there in two years.

Sources

  1. Donating the Model Context Protocol and establishing the Agentic AI Foundation, Anthropic, December 9, 2025.
  2. OpenAI launches AgentKit to help developers build and ship AI agents, TechCrunch, October 6, 2025.
  3. A new era of intelligence with Gemini 3, Google, November 18, 2025.