Model Context Protocol

Overview

Model Context Protocol (MCP) is a protocol for connecting AI models and agents to external tools, data sources, and servers in a standardized way. Instead of building one-off integrations for each model-tool pair, MCP provides a common interface that helps systems discover capabilities, invoke tools, and exchange context more efficiently.

For AI Product Managers, MCP matters because it sits at the intersection of agent UX, tool interoperability, and cost control. As more AI products rely on agents that need to access CRMs, internal knowledge, SaaS apps, or workflow tools, a protocol like MCP can reduce integration complexity while improving reliability and scalability. Recent newsletter mentions also highlight two practical themes: discovering MCP servers dynamically and reducing token usage in tool interactions.

Key Developments

• 2026-01-10: Phil Schmid introduced mcp-cli, an open-source CLI for dynamic discovery of Model Context Protocol servers. The tool was noted for reducing MCP token usage by 99% and improving AI agent tool interactions.
• 2026-03-24: Teun Rutten highlighted HubSpot’s developer ecosystem, especially its use of Model Context Protocol to let AI tools such as Cursor and Claude interact directly with HubSpot data.

Relevance to AI PMs

• Design more scalable agent integrations: MCP can help teams avoid bespoke integrations for every tool and model combination. For PMs, that means faster partner onboarding, cleaner platform architecture, and a clearer path to ecosystem expansion.
• Improve cost and latency performance: The mention of mcp-cli reducing token usage by 99% signals that protocol and discovery design can materially affect operating costs. PMs should evaluate how tool discovery, context packaging, and invocation patterns impact token spend and response speed.
• Enable better enterprise workflows: MCP is especially relevant when AI assistants need safe, structured access to systems like CRMs, internal apps, or productivity tools. PMs can use MCP-style patterns to define which actions agents can take, what data they can access, and how those interactions are governed.

Related

• HubSpot: Referenced as an example of a developer ecosystem using MCP to expose product data and functionality to AI tools.
• Cursor: An AI coding tool mentioned as a client that can interact with MCP-enabled systems.
• Claude: An AI assistant mentioned alongside Cursor as benefiting from MCP-based access to external data.
• mcp-cli: An open-source CLI focused on dynamic discovery of MCP servers and reducing token overhead.
• AI agent: MCP is highly relevant to agent architectures because agents often need standardized access to tools, APIs, and external context.