logo
Published on

The New AI Protocol

Authors
  • avatar
    Name
    Strategic Machines
    Twitter
stack

Igniting Intelligent Integrations

If simple is the more intelligent choice, according to Occam's razor, then why are LLMs so difficult to use? We don't mean spinning up a chat interface on OpenAI to scan for new recipe ideas. We do mean creating a production workflow for handling hundreds or thousands of customer calls about your products. Yet, here we are - wrestling with new integrations and maintenance requirements around databases, prompt engineering, context windows, process workflows and reliability issues. You get the picture. Where is the razor when we need it?

Enter Model Context Protocol (MCP), a new standard introduced by Anthropic in November 2024, designed specifically to bridge the gap between human intent, AI agents, and the tools they need - a much needed win for simplicity (and intelligent integrations!).

Since OpenAI released function calling in 2023, I’ve been thinking about what it would take to unlock an ecosystem of agent and tool use... It’s clear that there needs to be a standard interface for execution, data fetching, and tool calling.

— Yoko Li, Partner at Andreessen Horowitz

Li further highlights MCP’s foundational value by noting that "MCP is an open protocol that allows systems to provide context to AI models in a manner that’s generalizable across integrations. The protocol defines how the AI model can call external tools, fetch data, and interact with services."

In essence, MCP servers act as universal translators—seamlessly converting natural language instructions into precise actions across countless software tools and APIs.

From Fragmentation to Unification

Before MCP, companies hoping to leverage AI in production systems faced significant hurdles:

  • Technical Debt Nightmare: Every integration required custom code, authentication, and error handling.
  • Context Collapse: Vital context was easily lost between different systems.
  • Redundant Computation: AI models repeatedly solved identical problems in isolation.
  • Integration Bottlenecks: Incorporating new data sources could take weeks instead of minutes.

With MCP, these obstacles can be addressed. Companies can spin up MCP servers directly from existing product documentation or APIs, enabling powerful AI capabilities without custom integrations.

At Strategic Machines, we've noted that the deployment of LLMs for useful production applications was essentially a content management issue. We believe, based on our initial rounds of testing, that MCPs are a first step in the right direction to addressing that challeneg

Democratizing Software Power

Consider the current landscape of professional tools:

  • Want to create sophisticated 3D models? Prepare to memorize countless shortcuts in Blender.
  • Need to analyze complex data? Get ready for Tableau’s intricate visualization learning curve.
  • Want a professional website without coding? Ironically, mastering platforms like Bubble or Webflow is equally daunting.
  • Are you hoping to deploy the latest Voice Agents to help customers explore products, book rooms, make payments or reschedule delivery? The integrations are anything but simple.

MCP revolutionizes this paradigm. By providing context-aware, natural language interactions, MCP servers reduce time-to-deploy and eliminate steep learning curves. Anyone—from senior executives needing quick visualizations to small business owners crafting custom applications—can leverage sophisticated tools instantly.

Real-World Impact: Practical Examples

Here are tangible scenarios where MCP clearly outshines traditional API integrations:

1. Trip Planning Assistant

  • Traditional APIs: Separate, complex integrations for calendar, airlines, and email.
  • Using MCP: Your AI assistant checks calendars, books flights, and sends confirmations via MCP.

2. Intelligent IDE (Integrated Development Environment)

  • Traditional APIs: Manual integration with file systems, version control, and documentation.
  • Using MCP: IDEs automatically interact with development tools through a unified MCP protocol, delivering richer, context-aware suggestions.

3. Complex Data Analytics

  • Traditional APIs: Manually maintain database connections and data visualization tools.
  • Using MCP: An AI analytics platform autonomously discovers and interacts with multiple databases and visualizations through MCP.

The Emergence of MCP Ecosystems

The rapid adoption of MCP has spawned vibrant marketplaces and hosting solutions such as Mintlify’s mcpt, Smithery, and OpenTools. These platforms streamline server discovery, enabling developers to discover, share, and enhance MCP servers.

Thousands of MCP servers have emerged in mere months, predominantly in software delivery and IDEs. However, we anticipate a powerful wave of specialized MCP servers, dedicated to business and scientific domains, dramatically simplifying the consumption and management of tens of thousands of commerce APIs.

Furthermore, we expect future releases of MCP servers, with more sophisticated functions for security and observability, will be bundled directly with products—embedding documentation, warranties, training, and updates directly accessible by AI agents. This seamless integration from customer requirement to manufacturing floor will create genuinely intelligent, fully-connected business ecosystems.

As Yoko Li insightfully noted, MCP promises to turn every client into an “everything app,” eliminating friction and empowering unprecedented business agility (although we believe the 'everything' will be domain limited).

The MCP era has begun, and its profound impact on AI and intelligent software integration, usability, and business operations is only just becoming more simple. Stay tuned as we continue to uncover how MCP's 'razor edge'reshapes the landscape of intelligent technology.