Koala vs MicroAgent

Koala-AI lets you transform CART transcripts into personalized study tools!

Learn AI agents from the inside out. MicroAgent shows you the agent loop, tool calls, planning, memory, and multi-agent coordination — the parts every other AI app hides behind a chat box. This is not another chatbot. It is a 12-lesson visual course that breaks down how a real AI agent runtime is built, from a minimal loop with one tool to a full multi-agent team with task isolation. WHY MICROAGENT • See inside every step: messages, tool calls, plans, task state, memory, collaboration • Replayable demos: deterministic scenarios you can run again and again • Source code you can follow: each lesson maps to real Swift / Python code • 12 lessons that build on each other, not 50 random tips • Built for iPhone and iPad, fully offline-readable • One-time purchase, no subscription WHAT YOU WILL LEARN ① s01 The Agent Loop — a minimal agent with a single bash tool, end to end ② s02 Tools — read, write, edit, bash; how a tool registry actually works ③ s03 TodoWrite — make the agent plan before it acts ④ s04 Subagents — isolate work and avoid context pollution ⑤ s05 Skills — load knowledge on demand with SKILL.md ⑥ s06 Compact — keep the context window healthy with summarization ⑦ s07 Tasks — file-backed task system with dependencies ⑧ s08 Background Tasks — long-running work without blocking the loop ⑨ s09 Agent Teams — multi-agent coordination via JSONL inbox ⑩ s10 Team Protocols — shutdown, plan approval, request-response patterns ⑪ s11 Autonomous Agents — idle-loop teammates that auto-claim tasks ⑫ s12 Worktree + Task Isolation — parallel execution with shared task board WHO IT IS FOR ① iOS / Swift / full-stack developers ready to move from "use AI" to "build agents" ② AI and LLM self-learners who have read about ReAct, reflection, tool calling and want to see them run ③ Tech leads, PMs and indie builders who need a real mental model of agent architecture, not just prompt tricks HOW IT IS DIFFERENT • Visual breakdown of every step — not another chat interface. See the agent loop, planning, memory and coordination in action. • Learn from deterministic demos — run the same scenario repeatedly and understand each step. • Engineering patterns for real agent systems — not just prompting tricks. • Built-in code visualization — each lesson maps directly to runnable Swift and Python source. CORE CONCEPTS COVERED agent loop, tool calling, function calling, ReAct, planning, memory, RAG, MCP, subagents, skills, context compaction, multi-agent systems, team protocols, autonomous agents, task isolation, worktree, prompt engineering, LLM runtime, agent harness, replayable trajectory, eval, prompt injection safety. PRICING Start free with s01 to s03 plus the full course overview. Unlock full access to s04–s12 including planning, memory, background tasks, multi-agent teams, autonomous agents, and worktree isolation. No subscription, ever. LANGUAGES English · 简体中文 · 日本語 — fully localized course content and UI. Contact us: sectojoy@gmail.com Terms & Conditions: https://www.apple.com/legal/internet-services/itunes/dev/stdeula/ Privacy Policy: https://zelonai.com/app/microagent/privacy-policy