The core insight from the 2026 📝Model Context Protocol (MCP) State of the Union isn't "MCP is winning" — it's that the bottleneck for AI value creation has shifted from model capability to connectivity.
Models are good enough. What they can't do is talk to your five 📝Software as a Service (SaaS) tools, your shared drive, and your 📝CRM simultaneously. Whoever solves that plumbing problem captures the value layer.
This memo maps five investment signals from the talk that non-institutional investors should be tracking.
1. Agent-Native > API-First
The MCP lead called 📝REST-to-MCP wrappers "cringe." This is a direct signal: wrapping legacy APIs for agents produces bad results. Companies designing for agent interaction from the ground up will structurally outperform incumbents bolting on MCP support after the fact. The analog is mobile-native apps vs. responsive websites circa 2012 — same content, wildly different outcomes.
Look for: New SaaS entrants whose primary interface is agent-facing, not human-facing. Companies where the MCP server is the product, not an afterthought.
2. Identity Is the Chokepoint
The talk flagged "cross-app access" — log in once with your enterprise IdP, and all MCP servers authenticate automatically. This means the identity layer becomes the gatekeeper for the entire agent connectivity stack. Whoever owns that handshake controls which agents can reach which systems.
Look for: Auth/identity companies positioning themselves as the 📝Single Sign-On (SSO) layer for agent-to-service connections (not just human-to-service). This is a land-grab that favors existing identity players (📝Okta, Auth0) but could also produce new entrants purpose-built for agent auth.
3. The Knowledge-Worker Automation Wave Is Next
The talk explicitly frames 2025 as coding agents and 2026 as general agents — financial analysts, marketers, operations. Coding agents were the easy case: local, verifiable, sandboxed, developer sitting right there. Knowledge-worker agents are harder but vastly larger in market size. The constraint isn't intelligence, it's connectivity to the 5-10 SaaS tools each role depends on.
Look for: Horizontal agent platforms that solve multi-service orchestration for non-technical users. The "📝Zapier for agents" framing is reductive but directionally right. Also: vertical agents for specific roles (financial analysis, marketing ops) that ship pre-connected to the standard tool stack for that role.
4. "📝SEO for AI" Is Emerging
Server discovery via .well-known URLs means agents will crawl websites and ask "is there an MCP server here I can use instead of scraping?" This creates a new optimization surface — companies will compete to be discoverable and usable by agents, not just by humans in search results. First movers who make their services agent-discoverable will capture disproportionate agent traffic.
Look for: The tooling layer that helps businesses publish and optimize their MCP presence. This is early — the spec isn't even final — but the pattern mirrors early SEO: whoever helps businesses show up in the new discovery channel builds a durable business.
5. Developer Experience Compounds into Ecosystem Lock-In
📝FastMCP beat 📝Anthropic's own 📝SDK purely on developer experience. In a protocol land-grab, the tools with the best DX capture the most builders, which produces the most servers, which attracts the most clients. This is the classic platform 📝flywheel.
Look for: Dev-tools companies focused on MCP server creation, testing, deployment, and monitoring. The "📝Vercel for MCP servers" — hosting + DX + analytics in one package. 📝Cloudflare is already positioning here. Also watch whether FastMCP or similar community projects get acquired or spin into companies.
The Meta-Pattern
The biggest signal in this talk isn't any single feature — it's that MCP hit 110 million monthly downloads faster than 📝React. That's not a protocol anymore, that's infrastructure. And infrastructure adoption at that velocity means every company building on top of it is riding a wave, not creating one.
Non-institutional investors tend to be early on infrastructure shifts because they're closer to the developer community where adoption signals show up first. The advantage isn't capital — it's pattern recognition. If you've used MCP, you already know which of these layers are working and which are still broken. That's the edge.
This thesis emerged from an MCP State of the Union talk in April 2026. The speaker is a core protocol maintainer — the signals here aren't speculation, they're roadmap. When the person building the protocol tells you what's coming in June, that's alpha.
The investment framing is deliberately non-institutional. Big funds will index into "AI infrastructure" broadly. The edge for smaller investors is specificity: knowing that agent-native beats API-wrapper, that identity is the chokepoint, that DX compounds into lock-in. These are the kinds of structural insights that come from being a builder in the ecosystem, not just reading about it.