menu
{ "item_title" : "From APIs to MCP with Python", "item_author" : [" Piyush Doorwar "], "item_description" : "Your assistant can read the order. Can it be trusted to do something about it?Most guides to connecting AI models with business systems stop at the first successful tool call. This book starts there. You build one support assistant for a small online retailer in Python and take it all the way, joining Claude, a Shopify store, a Slack channel and the company's own operations service through the Model Context Protocol (MCP).The hard parts come after the demo works. Which employee was that call made for? What happens when the model proposes a refund? Who approved it? What does the system do when the network drops the reply, and how does anyone find out later what actually happened? Each chapter takes one of those questions and answers it in working code.What you will buildAn MCP server over Shopify's Admin API and an internal operations service, with tools that return a small set of checked factsA Python host that runs Claude through a tool loop with budgets and clear stopping rulesVerified employee identity on every tool call, checked against each support caseProposals that a second employee approves exactly as shown, and a worker that carries them out onceRecovery from lost replies through reconciliation and an outboxUnit tests, live evaluations against Claude, prompt-injection measurements, traces and a deployable stackBuilt and run for realEvery code listing comes from a public companion repository on GitHub, with one git tag per chapter. The project was run against a real Shopify development store, a real Slack workspace and the Anthropic API, and the outputs shown in the book were printed by that code.Who it is forPython developers who build integrations inside a company: support tools, operations tools, internal assistants. You should be comfortable with functions, classes, type hints and a little async, and you should have called an HTTP API before. You do not need a machine learning background or any prior experience with MCP.Covers MCP protocol revision 2026-07-28, FastMCP 4.1.4, the Anthropic Python SDK and the Shopify Admin API 2026-10. Appendices cover setup, just enough GraphQL, OAuth for MCP servers, connecting your own store and workspace, troubleshooting, and a chapter-by-chapter checkpoint map.", "item_img_path" : "https://covers3.booksamillion.com/covers/bam/9/79/816/667/9798166678522_b.jpg", "price_data" : { "retail_price" : "12.89", "online_price" : "12.89", "our_price" : "12.89", "club_price" : "12.89", "savings_pct" : "0", "savings_amt" : "0.00", "club_savings_pct" : "0", "club_savings_amt" : "0.00", "discount_pct" : "10", "store_price" : "" } }
From APIs to MCP with Python|Piyush Doorwar

From APIs to MCP with Python : Build a retail operations assistant with Claude, Shopify, and Slack

local_shippingShip to Me
In Stock.
FREE Shipping for Club Members help

Overview

Your assistant can read the order. Can it be trusted to do something about it?

Most guides to connecting AI models with business systems stop at the first successful tool call. This book starts there. You build one support assistant for a small online retailer in Python and take it all the way, joining Claude, a Shopify store, a Slack channel and the company's own operations service through the Model Context Protocol (MCP).

The hard parts come after the demo works. Which employee was that call made for? What happens when the model proposes a refund? Who approved it? What does the system do when the network drops the reply, and how does anyone find out later what actually happened? Each chapter takes one of those questions and answers it in working code.

What you will build

  • An MCP server over Shopify's Admin API and an internal operations service, with tools that return a small set of checked facts
  • A Python host that runs Claude through a tool loop with budgets and clear stopping rules
  • Verified employee identity on every tool call, checked against each support case
  • Proposals that a second employee approves exactly as shown, and a worker that carries them out once
  • Recovery from lost replies through reconciliation and an outbox
  • Unit tests, live evaluations against Claude, prompt-injection measurements, traces and a deployable stack

Built and run for real

Every code listing comes from a public companion repository on GitHub, with one git tag per chapter. The project was run against a real Shopify development store, a real Slack workspace and the Anthropic API, and the outputs shown in the book were printed by that code.

Who it is for

Python developers who build integrations inside a company: support tools, operations tools, internal assistants. You should be comfortable with functions, classes, type hints and a little async, and you should have called an HTTP API before. You do not need a machine learning background or any prior experience with MCP.

Covers MCP protocol revision 2026-07-28, FastMCP 4.1.4, the Anthropic Python SDK and the Shopify Admin API 2026-10. Appendices cover setup, just enough GraphQL, OAuth for MCP servers, connecting your own store and workspace, troubleshooting, and a chapter-by-chapter checkpoint map.

This item is Non-Returnable

Details

  • ISBN-13: 9798166678522
  • ISBN-10: 9798166678522
  • Publisher: Independently Published
  • Publish Date: October 2026
  • Dimensions: 9 x 6 x 0.56 inches
  • Shipping Weight: 0.79 pounds
  • Page Count: 266

Related Categories

You May Also Like...

    1

BAM Customer Reviews