Open reference library / 523 lessons
Study the mechanics. Build the system.
Explore the full English lesson sequence from AI Engineering from Scratch, alongside our governed agent field course. The original curriculum spans mathematics, machine learning, language models, tools, agents, infrastructure, safety, and capstones.
Reference material by Rohit Ghumare and contributors, MIT licensed. Imported at revision bf7791e14076. Runnable code and outputs remain linked to the source repository.
Connect your agent to this foundation ↗ — public search and lesson retrieval, with a reusable onboarding guide.
31 lessons
13 / tools and protocols The Tool Interface — Why Agents Need Structured I/O ~45 minutes ↗13 / tools and protocols Function Calling Deep Dive — OpenAI, Anthropic, Gemini ~75 minutes ↗13 / tools and protocols Parallel Tool Calls and Streaming with Tools ~75 minutes ↗13 / tools and protocols Structured Output — JSON Schema, Pydantic, Zod, Constrained Decoding ~75 minutes ↗13 / tools and protocols Tool Schema Design — Naming, Descriptions, Parameter Constraints ~45 minutes ↗13 / tools and protocols MCP Fundamentals: Stateless Requests and JSON-RPC ~55 minutes ↗13 / tools and protocols Building an MCP Server: Stateless Python and TypeScript ~85 minutes ↗13 / tools and protocols Building an MCP Client: Discovery, Routing, and Dual-Era Fallback ~85 minutes ↗13 / tools and protocols MCP Transports: stdio and Stateless Streamable HTTP ~65 minutes ↗13 / tools and protocols MCP Resources and Prompts: Addressable Context for Stateless Servers ~60 minutes ↗13 / tools and protocols MCP Model Input: Sampling Migration and Stateless MRTR ~75 minutes ↗13 / tools and protocols Explicit Scope and Stateless Elicitation ~60 minutes ↗13 / tools and protocols MCP Tasks Extension: Durable Work on a Stateless Core ~90 minutes ↗13 / tools and protocols MCP Apps on the Stateless Protocol ~75 minutes ↗13 / tools and protocols MCP Security: Poisoned Metadata, Routing, and MRTR State ~60 minutes ↗13 / tools and protocols MCP Authorization: CIMD, Issuer Binding, PKCE, and Step-Up ~90 minutes ↗13 / tools and protocols Stateless MCP Gateways and Registry Admission ~75 minutes ↗13 / tools and protocols MCP Auth in Production: Issuer-Bound Enrollment and Tokens ~90 minutes ↗13 / tools and protocols A2A — Agent-to-Agent Protocol ~75 minutes ↗13 / tools and protocols OpenTelemetry GenAI — Tracing Tool Calls End-to-End ~75 minutes ↗13 / tools and protocols LLM Routing Layer — LiteLLM, OpenRouter, Portkey ~45 minutes ↗13 / tools and protocols Agent Skills: Portable Contract and Runtime Boundary ~90 minutes ↗13 / tools and protocols Capstone: Stateless Tool Ecosystem ~120 minutes ↗13 / tools and protocols Skill Discovery and Progressive Disclosure ~105 minutes ↗13 / tools and protocols Skill Invocation and Routing ~105 minutes ↗13 / tools and protocols Skill Permissions, Sandboxes, and Trust ~120 minutes ↗13 / tools and protocols Skill Evals, Packaging, and Portability ~150 minutes ↗13 / tools and protocols MCP Tool Contracts and Content ~120 minutes ↗13 / tools and protocols MCP Reliability, Cancellation, and Flow Control ~120 minutes ↗13 / tools and protocols MCP Registry Supply Chain: Admission, Drift, and Rollback ~90 minutes ↗13 / tools and protocols MCP Conformance Engineering: Versioning, Evidence, and Operations ~100 minutes ↗