Tutorials

Each tutorial is a complete exercise with the commands to run and the output to expect. They build on each other loosely; the first one is the only prerequisite for the rest.

#	Tutorial	You will	Time
1	Your first task	Install malac, add a provider, fix a bug and review the result	10 min
2	Delegating to subagents	Run a research task and a refactor in parallel, and watch the agent tree in the terminal and the web UI	15 min
3	Automating a browser	Set up the control service and a browser profile, log in once, and scrape a page	20 min
4	Long-term memory with Milvus	Run Milvus in Docker, configure embeddings, save and recall memories	20 min
5	Running malac unattended	Use autopilot with deny rules, get --json output in CI, and review the session afterwards	20 min
6	Customizing malac	Write an agent type and a skill, add MCP servers, and configure malac by asking it	20 min

Before you start

• A terminal on macOS or Linux.
• An API key for a model provider, or a local model through Ollama. The examples use a gateway at lora.example.com with the models lora/default, lora/smart and lora/fast; use your own provider and model names.
• git and python3 for the example projects.

Model output varies between runs and models. Your transcripts will differ in wording and in the exact steps the agent takes; the shape of the output is what to look for.