Phase 13
Tools and Protocols
Phase 13: Tools and Protocols. 23 hands-on lessons building AI from first principles in the browser. Free reading; graded exercises and certificate with lifetime access.
- The Tool Interface — Why Agents Need Structured I/O (graded)
- Function Calling Deep Dive — OpenAI, Anthropic, Gemini (graded)
- Parallel Tool Calls and Streaming with Tools (graded)
- Structured Output — JSON Schema, Pydantic, Zod, Constrained Decoding (graded)
- Tool Schema Design — Naming, Descriptions, Parameter Constraints (graded)
- MCP Fundamentals — Primitives, Lifecycle, JSON-RPC Base (graded)
- Building an MCP Server — Python + TypeScript SDKs (graded)
- Building an MCP Client — Discovery, Invocation, Session Management (graded)
- MCP Transports — stdio vs Streamable HTTP vs SSE Migration (graded)
- MCP Resources and Prompts — Context Exposure Beyond Tools (graded)
- MCP Sampling — Server-Requested LLM Completions and Agent Loops (graded)
- Roots and Elicitation — Scoping and Mid-Flight User Input (graded)
- Async Tasks (SEP-1686) — Call-Now, Fetch-Later for Long-Running Work (graded)
- MCP Apps — Interactive UI Resources via `ui://` (graded)
- MCP Security I — Tool Poisoning, Rug Pulls, Cross-Server Shadowing (graded)
- MCP Security II — OAuth 2.1, Resource Indicators, Incremental Scopes (graded)
- MCP Gateways and Registries — Enterprise Control Planes (graded)
- MCP Auth in Production — DCR, JWKS Rotation, Audience-Pinned Tokens on iii Primitives (graded)
- A2A — Agent-to-Agent Protocol (graded)
- OpenTelemetry GenAI — Tracing Tool Calls End-to-End (graded)
- LLM Routing Layer — LiteLLM, OpenRouter, Portkey (graded)
- Skills and Agent SDKs — Anthropic Skills, AGENTS.md, OpenAI Apps SDK (graded)
- Capstone — Build a Complete Tool Ecosystem