Find your next step
ZettCode works in your terminal, on the files in your workspace. This guide is organized around what you want to do — start a first task, keep a conversation moving, or set up the tools and models that suit your project.
First time here?
- Install and connect a model. You need Python 3.10+, a terminal, and an OpenAI-compatible API. The quickstart takes you from installation to your first task.
- Get to know the screen. Understand thinking, tool output, the composer, and the status line.
- Keep the essentials handy. Enter submits, Ctrl-C stops a request, and Ctrl-D exits when the composer is empty.
Working on a project
| I want to… | Go to |
|---|---|
| Switch models, change reasoning effort, or ask a side question | Commands |
| Select text, paste an image, or scroll back to an answer | Keys & mouse |
| Continue an earlier session or export it to HTML | Sessions |
| Understand context usage, compaction, and caching | Models & context |
Typing while the agent works sends a steering message that it picks up at the next tool or model boundary. Use /btw <question> instead when you want an answer that will not join the context of later requests. See The interface for both.
Make it yours
- Configuration: model connections, optional capabilities, and your theme. Start here when you want to change a setting.
- Project instructions: write
AGENTS.mdto explain conventions, test commands, and boundaries for your workspace. - Skills & MCP: give the agent reusable instructions or connect external tools.
- Plugins: install Python extensions that add commands and customize the interface.
Something not working?
Start with Problems & questions. For a reproducible bug, open an issue on GitHub. Include the OS, terminal, ZettCode version, and the steps to reproduce it; leave API tokens and private conversation content out.