A guide to running multi-agent AI projects with the NBS framework.
You have Claude Code installed. You are comfortable with the terminal. You want to run projects where multiple AI agents collaborate on real work -- writing code, running tests, making architectural decisions -- with a human in the loop.
Chapters 1-3 get you running. Read them in order. By the end of Chapter 3 you will have launched a session and watched agents complete a task.
Chapters 4-6 explain how things work. Read them when you want to understand what you saw in Chapter 3, or when something does not behave as expected.
The Terminal Operator Tutorial is a standalone guide for operating the chat terminal — read it once your team is running.
Chapter 7 is a reference. Go there when something breaks.
Every chapter uses the same project: a C11 interpreter written in Python. The interpreter has four components -- lexer, parser, type checker, evaluator -- that map naturally to concurrent work streams. When the guide says "spawn a worker to implement the parser," it means this project.
- What NBS Is and Why -- The problem, the approach, the running example.
- Setup -- Install, verify, create your project directory.
- Running Your First Session -- Launch, observe, shut down. The hello world.
- Terminal Operator Tutorial -- Search, filter, control agents, spawn oracles, understand the display.
- How the Team Works -- Roles, the 3Ws, self-check, task scoping.
- Communication -- Chat commands, @mentions, the event bus, slash commands.
- Oracles -- Librarian, Pythia, Shepard, Fixup. Periodic assessment workers.
- Troubleshooting -- Common problems, diagnosis, fixes.