Skip to content

Tutorials

These tutorials walk you from a single boscli ask command to a packaged, shareable extension — each one building on the last. By the end you will have a working multi-agent project with custom tools, a persistent channel, and a Python package you can install in any BOS project.

Prerequisites

  • Python 3.13 or newer — BOS requires it; check with python3 --version.
  • An API key for at least one LLM provider (OpenAI, Anthropic, Gemini, DeepSeek, …). BOS uses LiteLLM under the hood, so most providers work out of the box with the right env var.
  • bos-ai installed — or use uvx boscli to run without installing. See Tutorial 1 for the exact install command.

New to BOS?

Start at Tutorial 1 and follow in order. Each page ends with a Next link and a summary of what you learned, so it's easy to pick up where you left off.

The tutorial track

# Tutorial What you learn
1 Your first agent Install BOS, set an API key, and run a one-shot query from the command line.
2 Create a project Scaffold a project with boscli init, understand the generated layout and config, and start the gateway + TUI.
3 Add a custom tool Drop a Python file into .bos/extensions/, decorate a function with @ep_tool, and have the agent call it.
4 Delegate to sub-agents Create specialist agents in Markdown files and wire them up with SubagentPlugin so the main agent can delegate tasks.
5 Connect a channel Attach a Telegram bot so users can talk to your agent from their phone.
6 Package & share extensions Turn your extensions into an installable Python package that registers itself automatically in any BOS project.

How the tutorials fit together

Tutorial 1 ──► Tutorial 2 ──► Tutorial 3 ──► Tutorial 4
  (install)      (project)      (tools)       (sub-agents)
                                          Tutorial 5 ──► Tutorial 6
                                          (channels)     (packaging)

Tutorials 3–6 all build on the project created in Tutorial 2. Keep the my-agent/ directory around as you work through the track.