@current-docs/cli

The command-line tool for Current: documentation that stays current with the code, by construction.

npx @current-docs/cli init

Self-contained: the engine, harness, standard checker, and the interactive wizard are bundled in, so the package has zero runtime dependencies and runs on plain Node (>=18), no Bun required.

Onboard a repo

current init runs an interactive wizard that detects your stack and scaffolds a working install.

• Harness: which coding-agent CLI runs your model work (Claude Code, Codex, Gemini CLI, pi, opencode, or Cursor).
• Docs mode: standalone site, embedded in a host app, or both.
• Routes: base path for the site, mount route for the embed.
• Brand: apply your app's shadcn / Tailwind colors, fonts, and logo to the docs.
• Components and APIs: which to render live and derive reference docs from.

It then writes the agent skills, a Fumadocs docs app, starter content, lefthook.yml, and current.config.json. Existing files are never overwritten.

current init             # interactive
current init --yes       # non-interactive, detected defaults, zero spend
current init --dry-run   # print the plan, write nothing
current init --bootstrap # also run the docs-setup agent for a first doc set

Daily use

| Command | Does | | ------- | ---- | | current loop enable | Arm the commit to docs-proposal loop (--auto-merge opt in) | | current lint | Check content against the authoring standard | | current check | Classify a diff for doc impact | | current update | Draft a docs proposal for staged changes | | current review [id] | List proposals, or show a diff and start a preview | | current merge <id> | Publish a proposal | | current drift | Fix docs affected by component or API changes | | current brand sync | Extract the host app brand into the docs | | current coverage [--fix] | Map code surface to docs pages; fill gaps | | current changelog <tag> | Build a release changelog page | | current stats | Loop metrics from telemetry | | current run <skill> | Run any skill through the harness | | current mcp | Serve docs to AI agents over stdio (MCP) |

Where the work runs

Generation runs on your own coding-agent subscription, on your machine. The platform runs no model and stores no code. init, lint, and stats are free; the loop, brand sync, and run spend on your agent.

See the main README for the full picture.