Model Context Protocol in Practice

Contents

12 chapters, about 5 hours 25 minutes to read or 7 hours 32 minutes to listen to, with 70 examples.

1. 1MCP end to end

   By the end of this chapter you can put an existing HTTP API behind a Model Context Protocol server over stdio, call its tool from a client, read every message the two exchange, and let a model choose the tool through a small host.

   21 minutes5 examples

   7 sections

   • Meet the shop and its API
   • Wrap one endpoint as a tool
   • Call the tool from a client
   • Read the messages on the wire
   • Let a model choose the tool
   • What goes wrong in production
   • Decisions this chapter made
2. 2What MCP adds to an API

   By the end of this chapter you can say what the host, the client and the server each do, put each of the shop's capabilities in the right Model Context Protocol primitive, read the version and capabilities that every request carries, and decide when the shop should keep a plain HTTP API instead.

   26 minutes6 examples

   11 sections

   • Separate the host, the client and the server
   • Put each capability in the right primitive
   • Tools are actions the model chooses
   • Resources are data the application chooses
   • Prompts are templates the user chooses
   • Carry the version and capabilities on every request
   • Handle a version the server doesn't support
   • Cache list results
   • Keep an HTTP API where MCP adds nothing
   • What goes wrong in production
   • Decisions this chapter made
3. 3Design tools a model can use

   By the end of this chapter you can turn the shop's API into a small set of Model Context Protocol tools whose names, descriptions, input schemas, results and errors a model can act on, and measure how a change to a description changes which tool a model picks.

   32 minutes7 examples

   10 sections

   • Shape tools around tasks, not endpoints
   • Write descriptions the model reads
   • Type inputs with JSON Schema
   • Return structured content with an output schema
   • Report failures as tool results
   • Mark what a tool changes with annotations
   • Treat annotations as hints, not guarantees
   • Keep the tool list small
   • What goes wrong in production
   • Decisions this chapter made
4. 4Serve data as resources and prompts

   By the end of this chapter you can publish the shop's policies and records as Model Context Protocol resources, address many records with one resource template, complete an argument while a user types it, tell a client when a resource changes, and package a refund review as a prompt the user can pick.

   25 minutes5 examples

   8 sections

   • Expose the policies as resources
   • Address records with resource templates
   • Complete arguments for the user
   • Tell clients when a resource changes
   • Package a workflow as a prompt
   • Choose between a resource and a tool
   • What goes wrong in production
   • Decisions this chapter made
5. 5Ask the user in the middle of a call

   By the end of this chapter you can make the shop's refund tool stop in the middle of a call and ask the person at the host to confirm a refund, with Model Context Protocol elicitation inside a Multi Round-Trip Request, on a server that keeps no call open while it waits for the answer.

   27 minutes6 examples

   9 sections

   • See why a stateless server can't wait for an answer
   • Return an input request instead of a result
   • Resume the call with the answer
   • Ask for structured input with elicitation
   • Send sensitive steps to a URL
   • Keep round-trip state tamper-proof
   • Choose elicitation or a tool argument
   • What goes wrong in production
   • Decisions this chapter made
6. 6Serve over Streamable HTTP

   By the end of this chapter you can serve the shop's Model Context Protocol server over Streamable HTTP, read what each request and response carries, stream progress from a long call and cancel it, run the server as replicas behind one address with no sticky sessions, and refuse requests that a browser was tricked into sending.

   28 minutes6 examples

   9 sections

   • Move the server from stdio to HTTP
   • Read one HTTP exchange
   • Stream progress on a long call
   • Cancel a call the client no longer needs
   • Run replicas without sticky sessions
   • Reject requests from other origins
   • Choose stdio or Streamable HTTP
   • What goes wrong in production
   • Decisions this chapter made
7. 7Run long work as tasks

   By the end of this chapter you can negotiate a Model Context Protocol extension, run the shop's slow restock report as a task that the client polls, let tasks expire and cancel them, and say what MCP Apps add to a tool.

   26 minutes4 examples

   8 sections

   • Negotiate an extension
   • Start a task instead of waiting
   • Poll a task and fetch its result
   • Expire and cancel tasks
   • Know what MCP Apps add
   • Choose a task, progress or a plain call
   • What goes wrong in production
   • Decisions this chapter made
8. 8Authorize clients with OAuth

   By the end of this chapter you can protect the shop's Model Context Protocol server with OAuth.

   31 minutes7 examples

   9 sections

   • Find the authorization server from the resource
   • Identify the client with a metadata document
   • Get a token with PKCE
   • Bind each token to one server
   • Validate the issuer of each response
   • Scope tools by what the token allows
   • Ask for more scope when a call needs it
   • What goes wrong in production
   • Decisions this chapter made
9. 9Secure tools against misuse

   By the end of this chapter you can draw the trust boundaries around a Model Context Protocol server and its host, refuse token passthrough, keep a server from acting as a confused deputy, pin tool definitions so a changed one is caught, treat ticket text as data, and ask the user before any tool call with side effects.

   27 minutes6 examples

   9 sections

   • Draw the trust boundaries
   • Never pass a client's token through
   • Keep the server from acting as a confused deputy
   • Catch poisoned tool descriptions
   • Pin tool definitions and flag changes
   • Treat ticket text as untrusted data
   • Ask for consent before side effects
   • What goes wrong in production
   • Decisions this chapter made
10. 10Build a host that works through tickets

    By the end of this chapter you can connect a host to several Model Context Protocol servers, give their tools names that can't collide, run a short loop in which a model calls tools and reads their results, answer a server's input requests on the user's behalf, put resources into the model's context, and work a queue of tickets through all of it.

    28 minutes6 examples

    8 sections

    • Connect one client per server
    • Name tools across servers
    • Run the host loop
    • Answer input requests for the user
    • Put resources in the model's context
    • Work through the ticket queue
    • What goes wrong in production
    • Decisions this chapter made
11. 11Test a server and watch it run

    By the end of this chapter you can test a Model Context Protocol server's tools in memory and over HTTP, catch a schema change that would break clients before it ships, replay a recorded host run as a regression test, trace tool calls with OpenTelemetry, measure how often a model chooses the right tool, and retire a tool without breaking the clients that use it.

    28 minutes6 examples

    8 sections

    • Test tools in memory
    • Check schemas as a contract
    • Replay recorded host runs
    • Trace calls with OpenTelemetry
    • Measure how often the right tool is chosen
    • Change tools without breaking clients
    • What goes wrong in production
    • Decisions this chapter made
12. 12Migrate a 2025-era server

    By the end of this chapter you can move a Model Context Protocol server written for the 2025 revisions and the Python SDK's 1.x line to the specification at revision 2026-07-28 and mcp 2.3.0: find what it depends on, remove its sessions, move it off the HTTP+SSE transport, replace sampling, roots, logging and Dynamic Client Registration, port the code, and serve clients of both eras while the old ones are still around.

    26 minutes6 examples

    9 sections

    • Inventory what the old server depends on
    • Remove initialize and session IDs
    • Move from HTTP+SSE to Streamable HTTP
    • Replace sampling, roots and logging
    • Replace dynamic client registration
    • Port the code from SDK 1.x to 2.x
    • Serve old and new clients through the window
    • What goes wrong in production
    • Decisions this chapter made