Redistributed

General

Model Context Protocol (MCP)

Published March 8, 2026

standardsinfrastructuretech

An open standard for connecting AI assistants to external tools and data sources — a bet that interoperability, not lock-in, should define AI infrastructure.

What It Is

MCP is a protocol specification that standardizes how AI models communicate with external systems — databases, APIs, file systems, and other tools. Think of it like USB for AI: a common interface that any tool can implement.

Developed by Anthropic and released as an open specification, MCP enables:

  • Tool interoperability — Write a tool once, use it with any MCP-compatible AI
  • Context sharing — AI assistants can access the same data sources
  • Composability — Build complex agent systems from standardized components

Why It Matters for Redistribution

Who benefits:

  • Developers building AI tools — no more proprietary integrations per platform
  • Users — tool investments aren't locked to one AI provider
  • Smaller AI labs — can compete on model quality, not ecosystem size

Who loses power:

  • Platform incumbents who benefit from ecosystem lock-in
  • Companies using proprietary protocols to capture users

What changes: Open standards reduce switching costs. If your tools work with any AI, you're not locked to whoever you started with. This matters as AI becomes infrastructure.

The tension: Anthropic developed MCP, so there's a question of governance. Will it become a true open standard with community control, or remain shaped by its originator? The protocol is open, but the governance is still evolving.

The redistribution stakes are high. As Compute Is the New Oil argues, control points in AI infrastructure determine who captures value. Open standards reduce the power of those control points — much as HTTP reduced the power of proprietary networking protocols. But the history of open standards also shows that originating companies can shape standards to their advantage even when the specification is technically open. The question explored in Open (For Business) applies here: is MCP open because interoperability genuinely serves users, or because Anthropic benefits from an ecosystem built around its protocol?

MCP also matters for How Technology Gets Distributed. Standardized tool interfaces mean that AI capabilities can spread more evenly — a small developer building an MCP server gets the same integration surface as a large platform. This is the kind of infrastructure leveling that redistribution requires.

Current Status

Emerging — MCP launched in late 2024 and is gaining adoption. Claude Code, Cursor, and other tools support it. The specification is open, with reference implementations in Python and TypeScript.

Key milestones:

  • Specification published and versioned
  • Reference servers for common use cases (filesystem, databases, web)
  • Growing ecosystem of community-built tools
  • Questions about governance and evolution remain open

How to Get Involved

  • Use it: Build MCP servers for your tools and data sources
  • Contribute: Submit improvements to the specification
  • Advocate: Push for MCP support in AI tools you use
  • Watch governance: Pay attention to how the standard evolves

Sources

← Back to all articles
On this page