MCP S‑Curve Building
Michael Hochstat predicts MCPs (Model Context Protocol) will follow an S-curve — exploding developer adoption first, slower enterprise uptake — with MCPs acting as the 'HTTP of AI' for agents and tool access predicted. The key selling point is governance and auditability: MCPs let enterprises lock policies around what agents can call and where context lives.
[Anthropic announced]anthropic.com the Model Context Protocol on Nov 25, 2024 and published the initial specification, SDKs, and a repository of reference MCP servers as part of that release. Major platform vendors began documenting MCP support — OpenAI’s Agents [SDK includes]openai.github.io MCP guidance and IBM published MCP integration and AgentOps [tutorials describing]ibm.com enterprise patterns for connecting MCP servers to business systems. The MCP project’s spec repository shows active community maintenance with thousands of commits and roughly 7.4k stars on [GitHub evidencing]github.com growing developer investment, and the project published an updated roadmap on Mar 9, 2026 prioritizing transport scalability, agent communication, and enterprise [governance details]blog.modelcontextprotocol.io. Enterprise momentum and risk surfaced in parallel: Microsoft published a Feb 12, 2026 blog on using MCP inside Microsoft 365 Copilot while stressing governance [controls announced]microsoft.com; independent research analyzing 1,899 open-source MCP servers found maintainability issues and measured 7.2% general vulnerabilities and 5.5% MCP-specific tool‑poisoning [incidents reported]arxiv.org; meanwhile vendors such as [Bluente released]freep.com an open-source MCP server and Microsoft shipped a C# MCP SDK update last [week noted]infoworld.com.
