How I Think About MCP: A Practical Guide to Tool Use in AI Agents

Source: Nate’s Newsletter Date: 2025-04-09 URL: https://natesnewsletter.substack.com/p/how-i-think-about-mcp-a-practical

Summary

Paywalled — the visible premise frames MCP as a paradigm shift toward “giving AI the power to actually do things” rather than explicit instruction. Nate covers where MCP does and doesn’t work, comparing it to APIs and LangChain as alternatives.

Implications

Agent product strategy thread. MCP vs. API vs. LangChain as a practical tool-use decision framework is exactly the architectural choice that agent builders face in 2025. Understanding Nate’s practical mental model for when each approach fits is operationally valuable for teams building agent toolchains.

Vendor positioning thread. The comparison to LangChain positions MCP as Anthropic’s answer to the framework abstraction problem — giving AI tool access through a protocol rather than a library. Whether MCP’s open-protocol design beats LangChain’s ecosystem depends on community adoption velocity.

Watch: Whether MCP adoption accelerates beyond Anthropic’s own ecosystem to become the dominant tool-integration protocol across multiple agent platforms and frameworks.