Developer guide
This chapter explains the architecture, implementation, and verification needed to develop Agent Board. If you only want to use the board, read the guide instead.
Agent Board is a single Rust program, agboard, that is both a CLI and a web server. All data is stored in files; the CLI, direct Markdown edits, and the browser work with the same data:
How this chapter is organized
| Part | What you will find | Read it when |
|---|---|---|
| Get started | Set up and run, follow a first change, use Storybook, verify and open a pull request. | You want to build and change something. |
| Concepts | Short pages that explain one idea each: the overall design, tasks and states, history and decisions, saving and recovery, notifications, the server, agents, the mobile monitor, and security. | You want to understand how it works before you change it. |
| Reference | Facts to look up: shared files, the web API, error codes, the JSON that agents receive, monitor routes, and the code map. | You know what you are looking for. |
| Troubleshooting | Symptoms, causes, and what to do. | Something fails. |
| Documentation and images | Previewing the site, UI Stories, and regenerating screenshots and videos. | You change docs/ or the UI. |
A path for your first day
- Set up and run: build it and run it against a temporary board.
- How Agent Board works: read the overview and follow one change through the system.
- Follow your first change: trace one rule from the CLI to the file and to its test.
- Tasks and states, then the concept page of the area you will touch.
- Verify and open a pull request.
Keeping these pages current
- They describe how Agent Board works now. Plans and the history of decisions are kept outside the repository.
- Update them in the same change as the behavior they describe.
- English is the source. The Japanese pages in
docs/ja/dev/are translations for developers; update them in the same change. - The repository's
AGENTS.mdindexes these pages for coding agents.