Awesome AI Agents › Courses and Tutorials
How Claude Code Works is a study project that documents the internal architecture of Claude Code as an agentic coding system. The authors state that it is an independent educational analysis for developers, that it does not represent official Anthropic design, and that no Anthropic source code is redistributed. The material was produced by reading a large leaked TypeScript snapshot together with the agent itself and distilling the findings into topic documents, published both in the repository and on a documentation site. Topics include an overview of technology choices and core design principles along with a nine-stage parallel startup path, the two-layer agent loop with seven continue sites for failure recovery, tool pre-execution that hides tool latency inside the model generation window, and a streaming tool executor for concurrency. Context engineering is covered in detail, including a four-level progressive compression pipeline of trimming, deduplication, folding and summarization, automatic restoration of recently edited files after compaction, and prompt caching strategy with cache-break detection. The tool system chapter describes registration and concurrency control, where read-only tools run in parallel and writes are serialized, oversized tool output is spilled to disk, and third-party MCP tools follow the same interface and safety checks as built-ins, with six MCP transports and an OAuth 2.0 with PKCE flow. The skills system chapter covers six skill sources and priorities, lazy loading with token budgets, inline and fork execution, and an allowlist permission model. Security is described as layered defense spanning workspace trust, permission modes, rule matching, syntax-tree analysis of shell commands with static checks, per-tool validation, process sandboxing with Git worktree isolation, and user confirmation. Multi-agent modes covering subagents, a coordinator and swarm communication are also documented. A companion project provides a from-scratch tutorial implementation.
https://github.com/Windy3f3f3f3f/how-claude-code-works
A comprehensive guide and resource hub for prompt engineering, context engineering, retrieval-augmented generation, and AI agents, offering tutorials, papers, tools, and courses to optimize the use of large language models.
Learn Claude Code is an educational project that teaches how to build modern AI coding agents like Claude Code through a progressive tutorial covering core concepts, planning, subagents, and skills mechanisms.
AI Engineering Hub is a comprehensive repository offering in-depth tutorials and practical resources on Large Language Models, Retrieval-Augmented Generation, and real-world AI agent applications for learners and practitioners.
The Hugging Face Agents Course is a comprehensive educational resource that teaches the fundamentals and advanced concepts of AI agents and large language models through structured units, practical frameworks, and community collaboration.
GenAI_Agents is a comprehensive repository offering tutorials, implementations, and community resources for building and advancing generative AI agents from basic to advanced levels.
This course provides a comprehensive project-based tutorial on Harness Engineering for building reliable AI coding agents, covering environment setup, state management, verification, and control.
An AI-powered tool that analyzes GitHub repositories and generates beginner-friendly tutorials with visualizations to simplify understanding complex codebases.
A comprehensive guide to prepare for machine learning and AI technical interviews at major tech companies, covering coding, ML fundamentals, system design, and behavioral aspects.