Analysis
Anthropic expanded Claude's support for the newly published MCP 2026-07-28 specification, the latest revision to the Model Context Protocol that Anthropic open-sourced as a standard for connecting AI models and agents to external tools and data sources. The update introduces a stateless core architecture, stronger OAuth and OIDC-based authorization, and versioned extensions covering enterprise use cases like embedded UI, enterprise-managed authentication, observability, and private network tunnels.
The enterprise-security focus of this revision is the notable shift. Earlier MCP versions prioritized developer ergonomics and quick tool integration; this spec update is explicitly aimed at the authentication, auditability, and network-isolation requirements that enterprise security and compliance teams have been citing as blockers to broader agent-connectivity adoption. That's a maturity signal for the protocol as a category-defining standard rather than just a convenient integration layer.
“The enterprise-security focus of this revision is the notable shift.”
MCP's adoption curve matters beyond Anthropic's own ecosystem: OpenAI, Google, and a wide range of tooling vendors have adopted or built compatibility layers for the protocol since its release, making it a rare instance of cross-lab technical standardization in a market otherwise defined by competing proprietary ecosystems. For infrastructure and dev-tools founders, a maturing MCP spec with enterprise-grade auth lowers the barrier to building agent-connectivity products that work across multiple frontier labs' agents rather than locking into a single vendor's proprietary tool-calling format -- a meaningful de-risking of a category that looked more fragmented a year ago.
