← Back to curriculum
Phase 13 · 23 lessons
Tools & Protocols
The interfaces between AI and the real world.
01The Tool Interface — Why Agents Need Structured I/OLearn02Function Calling Deep Dive — OpenAI, Anthropic, GeminiBuild03Parallel Tool Calls and Streaming with ToolsBuild04Structured Output — JSON Schema, Pydantic, Zod, Constrained DecodingBuild05Tool Schema Design — Naming, Descriptions, Parameter ConstraintsLearn06MCP Fundamentals — Primitives, Lifecycle, JSON-RPC BaseLearn07Building an MCP Server — Python + TypeScript SDKsBuild08Building an MCP Client — Discovery, Invocation, Session ManagementBuild09MCP Transports — stdio vs Streamable HTTP vs SSE MigrationLearn10MCP Resources and Prompts — Context Exposure Beyond ToolsBuild11MCP Sampling — Server-Requested LLM Completions and Agent LoopsBuild12Roots and Elicitation — Scoping and Mid-Flight User InputBuild13Async Tasks (SEP-1686) — Call-Now, Fetch-Later for Long-Running WorkBuild14MCP Apps — Interactive UI Resources via `ui://`Build15MCP Security I — Tool Poisoning, Rug Pulls, Cross-Server ShadowingLearn16MCP Security II — OAuth 2.1, Resource Indicators, Incremental ScopesBuild17MCP Gateways and Registries — Enterprise Control PlanesLearn18MCP Auth in Production — Enrollment, JWKS Refresh, Audience-Pinned TokensBuild19A2A — Agent-to-Agent ProtocolBuild20OpenTelemetry GenAI — Tracing Tool Calls End-to-EndBuild21LLM Routing Layer — LiteLLM, OpenRouter, PortkeyLearn22Skills and Agent SDKs — Anthropic Skills, AGENTS.md, OpenAI Apps SDKLearn23Capstone — Build a Complete Tool EcosystemBuild