Agent Handbook

Agent Handbook is a hands-on field guide for building AI agents — right from your iPhone. EIGHT BOOKS, 52 CHAPTERS - Foundations: prompts, context, tokens, latency - Reasoning models: when to use thinking vs non-thinking - Multi-modal: vision, audio, document understanding - Tools and MCP: how agents call the outside world - Multi-agent: coordinator, specialist, debate patterns - Streaming: tokens, deltas, partial tool calls - Vision: OCR, layout, screenshots - Production: cost, eval, observability, safety FIVE LIVE PLAYGROUNDS - Chat: talk to any model with streaming and chain-of-thought - AI Advisors: switch between specialist personas - Tool Playground: 12 simulated tools (pwd, list_directory, http_request, translate, run_shell and more) - MCP Simulator: explore the Model Context Protocol offline - Multi-agent: orchestrate two or more models in one conversation BRING YOUR OWN KEYS - OpenAI, Anthropic, Google, DeepSeek, plus any OpenAI-compatible endpoint - Curated 2026 model registry: GPT-5.5, Claude Opus 4.7, Gemini 3 Pro, DeepSeek V4 Pro - Capability badges (thinking, vision, audio, tools) on every model - Custom model tags so you can register your fine-tuned variants PRIVACY FIRST - Your API keys live in iOS Keychain. They never reach our servers - Direct HTTPS calls from your device to the provider you chose - Zero analytics, zero tracking, zero ads 6 LANGUAGES English, simplified Chinese, traditional Chinese, Japanese, Korean, Vietnamese.