Used for feedback · quests · score card
Build an AI agent from scratch, step by step. Each module builds on the previous one — by the end you'll have a complete AI agent connected to real-world tools.
Understand the big picture: what AI agents are, why they need tools, and how MCP connects them.
Install Docker, get API keys, clone the repo, and verify everything works.
Poke around the existing system with curl. See how tools/list and tools/call work.
Create an MCP-compliant server from scratch with JSON-RPC 2.0, tool discovery, and a weather tool.
Build an AI agent that discovers tools dynamically, talks to an LLM, and orchestrates tool calls.
Connect your agent and MCP server with Docker Compose and run the full system end-to-end.
Add a random facts tool to the MCP server. The agent discovers it automatically.
Add a news/RSS tool that pulls real data from external APIs.
Swap OpenAI for a local LLM using Ollama. No API key needed, runs on your machine.
Add error handling, debugging, context management, and production hardening.