TinkerChapter 5 of 7
MCP: one plug for many tools
Larkspur’s tools now live in two systems: orders and the team calendar. Other teams want AI apps of their own that use the same tools. Wiring every app to every system by hand will not last.
This chapter: How can any AI app use any system’s tools without a custom integration for each pair?
You will learn
- Why custom integrations multiply, and how one shared protocol turns the product into a sum
- Who is who in MCP: the host, its clients and the servers
- What a server offers besides tools: resources and prompts
Words you’ll meet (7)
- Integration
- Code that joins one AI app to one system so the app can use its tools.
- MCP
- The Model Context Protocol: an open standard for how AI apps talk to systems that offer tools and data.
- MCP server
- A program that offers one system’s tools, data and prompts to any app that speaks MCP.
- MCP client
- The connector inside a host app that talks to one server.
- JSON-RPC
- A simple message format: each request names a method, and gets back a result or an error.
- Resource
- Data a server offers as context, such as a policy document.
- Prompt
- A ready-made message template a server offers, for a person to pick.
The picture on the left is live. Each step below asks you to do one thing in it; the story continues here as you go.
Step 1 of 7
Larkspur’s tools now live in 2 systems: orders and the team calendar. Every AI app that wants a system’s tools needs code that joins the two. Other teams want apps of their own. Step up to 3 apps and 4 systems and watch the wiring.
AI apps
Systems with tools
2without MCP: 1 apps × 2 systems