GitHub radar
LearnHarness: an AI agent that teaches you at your own level
LearnHarness is an app where an AI agent figures out what you already know and guides you step by step toward your goal. Once installed, you get a visible lesson plan instead of generic explanations.
It's an AI agent (a self-directed assistant) that first tests your knowledge with questions, then builds and runs a personal lesson.
When it helps
Useful when you want to understand a specific topic through dialogue instead of reading a generic article. Not needed if a ready-made course is enough, or you have no API key or subscription.
Pros
- Adapts to your level instead of giving a generic explanation
- Shows you a visible plan toward your goal
- Stores knowledge in a graph and reuses it in later topics
- Works with Claude Code CLI instead of a browser chat
Cons
- Doesn't work without your own API key or app subscription
- Voice input and web search need separate service keys
How to set it up — step by step
- 1Download LearnHarness from github.com/amosblomqvist/learnharness or from learnharness.ai.
- 2Open the README in the repository and install the app for your system (macOS, Windows, Linux).
- 3Connect a model: add your API key, subscribe in the app, or install the Claude Code CLI.
- 4Send the agent the text from the block below, naming the topic you want to learn.
- 5Check the lesson plan the agent shows you and go through the first explanation step.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Download LearnHarness from https://github.com/amosblomqvist/learnharness (or learnharness.ai) — set it up per the README, connect a model (ask me which key or subscription to use), start a session on a topic I want to understand, and show me the lesson plan and the first explanation step.
Runs on a regular laptop — there are builds for macOS (Apple Silicon), Windows and Linux. It has no model of its own; the real load sits with whatever model you connect through your key or subscription.
Open on GitHub▌ More finds