Getting Started with Rapid OS
Rapid OS governs work by your coding agent: every change happens within a run with a spec, tasks, gates, and verifiable evidence. You do not need to learn those commands; the agent runs them through shortcuts (/rapid-...). In your terminal, you only install and approve.
1. Start in three steps
You need Python 3.10+, Git, and a coding agent: Claude Code, Codex, Cursor, Antigravity, or VS Code (Copilot).
Step 1: Install
Use the installer for your system (Linux, macOS, Windows, or WSL). At the end, it asks which agents should receive shortcuts; enter yours (for example, claude) to make them available across projects. If you skipped that step, run this once:
rapid shortcuts install --global --harness claude
VS Code has no user directory for shortcuts, so install them in each project with rapid init --shortcuts. Do the same if your Rapid OS version does not yet include /rapid-init or --global (check rapid shortcuts install --help).
Step 2: Describe the project with /rapid-init
Open your agent in the project folder (new or existing) and enter:
/rapid-init a Python API for validating usernames
The agent proposes the stack, test command, and project documents (brief, business rules, data model, and architecture), one decision at a time. For each document, you can provide it, have the agent draft it, or skip it. Agent-written content remains a draft until you confirm it. In Codex, enter $rapid-init.
Step 3: Request the work
/rapid-feature add CSV export to the monthly report
The agent drafts the spec, implements it task by task, and runs the tests. When a person needs to act, it stops and gives you the exact command for your terminal, for example:
rapid approve gate.final-verification
Then enter /rapid-next in the agent to close the run. No shortcut approves anything on your behalf. See When the Agent Stops for what to do at each pause.
| You want to... | Enter in your agent |
|---|---|
| Build a new feature | /rapid-feature <goal> |
| Fix a bug | /rapid-bugfix <what is broken> |
| Refactor without breaking tests | /rapid-refactor <scope> |
| Harden security | /rapid-hardening <area> |
| Investigate an option | /rapid-research <question> |
| Check progress or continue | /rapid-status · /rapid-next |
| See what needs approval | /rapid-approve |
See the shortcuts guide for shortcut locations and invocation in each agent.
2. Section map
Read these pages in order; each takes a few minutes.
| Page | What you will learn |
|---|---|
| 1. Your First Project | Go from an empty folder to your first approved feature. |
| 2. An Existing Project | Use Rapid OS in a coded repository without changing business code during setup. |
| 3. When the Agent Stops | Approve, reject, check progress, cancel, and read the verdict. |
| 4. Writing the Spec | What the agent asks to draft a spec and how to make a good request. |
| 5. What the Agent Knows About Your Project | Where its rules come from, drafts versus confirmed content, and how to change them. |
| 6. Next Steps | Labs, choosing a mode, teamwork, and troubleshooting. |
Also see the shortcuts guide, hands-on labs, CLI reference, and diagnostics.
Advanced mode
You do not need this to use Rapid OS. The v3 commands (rapid spec, rapid run, rapid evidence, rapid eval, and others) let you do manually what the agent does with shortcuts; they are useful for CI, debugging, and auditing. See How It Works:
- v3 Governance Loop: What each step records and why.
- Manual Walkthrough: A complete project done manually, from
rapid initto the verdict.