Continue Learning from the Source

Phone: 650.723.2465
Fax: 650.725.0253

MCP changes quickly. For protocol details, interface labels, and current capabilities, prefer the official documentation over any tutorial, including this one.

What was tested, and when. The exercises were built and verified against MCP Inspector 2.7.0, the MCP TypeScript SDK 1.30.0, and Zod 4.6.5, on Node.js 22.19.0 or newer. That combination negotiates an earlier protocol version than the latest published specification, which is normal while implementations catch up.

The Stanford Claude, Stanford ChatGPT Edu, and Claude Code interface observations in Modules 7 to 9 were verified on 17 September 2026. These are managed commercial products, and their labels, controls, and institutional configuration may change. Where this tutorial describes an interface, treat it as what we saw on that date rather than a permanent fact.

Model Context Protocol documentation (opens in a new tab) — the official home for the protocol and its tooling.
The current specification (opens in a new tab) — the authoritative definition of what clients and servers must do.
MCP Inspector (opens in a new tab) — official documentation for the inspection tool used in Module 1.
TypeScript SDK (opens in a new tab) — the SDK the practice server is built with.
Approved Connectors for AI Services (opens in a new tab) — which connectors Stanford has documented for each of its AI services.
Claude for Education at Stanford (opens in a new tab) and ChatGPT Edu at Stanford (opens in a new tab) — the two managed environments Module 7 inspects, including which data classifications each is approved for.
Responsible AI at Stanford (opens in a new tab), Responsible Agentic AI (opens in a new tab), and Use of Generative AI Technology (opens in a new tab) — the current Stanford and SLS guidance.

Back to the tutorials →