The Hitchhiker’s Guide to LLM Agents
Source: https://saurabhalone.com/hitchhikers-guide-llm-agent Author: Saurabh Alone Date: 2026-01-06
Summary
Comprehensive beginner-to-intermediate guide to LLM agents. Covers the full stack from basic agent loop through tool use, memory, and multi-agent systems. One of the better single-source overviews.
Key Claims
- Agent fundamentals: loop + tools + memory. Each component can be simple or complex independently.
- Tool categories: read (retrieval, search), write (file, API, browser), compute (code execution, calculations).
- Memory types: in-context (working), external (RAG/files), procedural (system prompt), episodic (conversation history).
- Multi-agent patterns: orchestrator/worker, peer-to-peer, hierarchical. Each has different coordination costs.
- Failure modes: hallucination, tool misuse, loop traps, context overflow.
Concepts
- Coding Agents — comprehensive agent architecture overview
- Agent Memory — memory taxonomy
- Context Engineering — context management strategies