Build your own AI coding agent from scratch. Learning the patterns behind Cursor, Aider, and Cline. Tool use, agentic loops, context management, and more.

Build Your Own AI Coding Agent (TypeScript)

A step-by-step workshop for building an AI-powered coding assistant using TypeScript and the Anthropic Claude API.

What You'll Learn

By the end of this workshop, you'll understand how to:

• Connect to the Anthropic Claude API
• Build a simple AI chatbot with conversation history
• Add tools like reading files and listing directories
• Handle tool requests and errors gracefully
• Build a modular agent framework that's easy to extend

What We're Building

You'll build 7 versions of a coding assistant, each adding more capabilities:

1. Basic Chat — talk to Cursor
2. File Reader — read code files
3. File Explorer — list files in folders
4. Modular Framework — clean, extensible architecture
5. Refined Toolkit — powerful tools like bash, edit, and grep
6. Streaming Responses — real-time text and tool visibility
7. Extended Thinking — see Cursor's reasoning process in real-time

graph LR
    subgraph "Application Progression"
        A["Chapter 1<br/>Basic Chat"] --> B["Chapter 2<br/>+ File Reading"]
        B --> C["Chapter 3<br/>+ Directory Listing"]
        C --> D["Chapter 4<br/>Modular Framework"]
        D --> E["Chapter 5<br/>Refined Toolkit"]
        E --> F["Chapter 6<br/>Streaming"]
        F --> G["Chapter 7<br/>Extended Thinking"]
    end

    subgraph "Tool Capabilities"
        H["No Tools"] --> I["read_file"]
        I --> J["read_file<br/>list_files"]
        J --> K["read_file<br/>list_files<br/>(modular)"]
        K --> L["bash, edit_file,<br/>grep (ripgrep)"]
        L --> M["+ streaming<br/>+ tool visibility"]
        M --> N["+ thinking blocks<br/>+ reasoning visibility"]
    end

    A -.-> H
    B -.-> I
    C -.-> J
    D -.-> K
    E -.-> L
    F -.-> M
    G -.-> N

How It Works (Architecture)

Each agent works like this:

1. Waits for your input
2. Sends it to Cursor
3. Cursor may respond directly or ask to use a tool
4. The agent runs the tool