Commands

Every command accepts --help for inline documentation.

Command summary

clawnify init [dir]

Scaffolds a new Clawnify project in dir (or the current directory if omitted). Asks whether you’re building an app (React + Hono + D1) or a website (Astro, public, one per org). Skip the prompt with --type app or --type website.

The app scaffold is an opinionated TypeScript project — typed backend, lightweight frontend, SQLite database — already wired for Clawnify’s build pipeline.

What ends up in the directory:

my-app/
  clawnify.json           # App manifest — name, framework, database
  package.json            # Dependencies + dev scripts
  tsconfig.json
  index.html              # Frontend entry
  src/
    server/
      index.ts            # API entry point
      routes.ts           # Procedures
      db.ts               # Database adapter
      schema.sql          # Applied on first deploy
    client/
      main.tsx
      app.tsx             # Root component

Plus a couple of local-only config files (Vite + dev server) that the CLI manages — you don’t need to edit them.

The website scaffold is a minimal Astro project (Tailwind v4, a BaseLayout, and a single index.astro). Its clawnify.json carries { "website": { "framework": "astro" } } instead of an app block — that’s the cue for clawnify deploy to publish it as a website.

After scaffolding, init runs clawnify ai-files install automatically so the project is agent-ready out of the box. Pass --no-ai-files to skip it.

clawnify deploy [dir]

Deploys dir (or the current directory) to production. The CLI reads clawnify.json and dispatches based on whether it declares an app or a website:

clawnify deploy                    # deploy current directory (app or website)
clawnify deploy ./my-app           # deploy a specific directory
clawnify deploy --from owner/repo  # deploy from a GitHub repository (apps only)

App projects ship to https://{slug}.apps.clawnify.com. Pending schema migrations are applied automatically as part of the build.

Website projects ship to https://{slug}.myclawnify.com. Each org has one website; the first deploy provisions it, every subsequent deploy overwrites the source.

The CLI streams build progress and prints the live URL when the build finishes.

clawnify ls

Lists every app in your active organization.

clawnify ls

clawnify open <slug>

Opens the deployed app in your browser with an authentication token attached to the URL so protected routes work immediately.

clawnify open my-app

clawnify logs <app-id>

Fetches the build status and log output for a given app.

clawnify logs <app-id>

clawnify rm <app-id>

Deletes an app and all of its resources (the deployed worker, its database, and source files).

clawnify rm <app-id>

clawnify whoami

Prints the currently logged-in user’s email and the active organization.

clawnify whoami

clawnify ai-files install

Makes a project agent-ready in one step. Installs into every coding agent detected in the directory (Claude Code, Cursor, Codex, OpenClaw, Aider) and defaults to Claude (CLAUDE.md) when none is found.

clawnify ai-files install
clawnify ai-files install --agent cursor   # force a specific target
clawnify ai-files install --no-mcp         # skip writing .mcp.json

It writes three things:

1. Rules — an always-on Clawnify guidance block in the agent’s rules file (CLAUDE.md / AGENTS.md / .cursorrules). Marker-delimited and version-pinned, so reruns replace it in place rather than appending duplicates.
2. Skills — the bundled skills (e.g. build-a-clawnify-app) into .claude/skills/, .cursor/rules/, etc.
3. MCP — a project .mcp.json wiring the Clawnify MCP server (idempotent merge that preserves any foreign servers).

clawnify init runs this automatically after scaffolding (opt out with --no-ai-files). Re-run it after a CLI upgrade to refresh rules and skills in place.

clawnify ai-files uninstall

Removes the Clawnify rules block, skills, and MCP server entry — leaving any surrounding human-written content and foreign MCP servers untouched.

clawnify ai-files uninstall
clawnify ai-files uninstall --agent cursor   # restrict to one agent

clawnify execute [file]

Run TypeScript on your agent’s runtime via mcp.clawnify.com. Streams stdout / stderr through; exits with the runtime’s exit code.

clawnify execute path/to/script.ts
clawnify execute -c "console.log(1 + 1)"
clawnify execute -c "..." --timeout 60

clawnify docs [<slug>]

Fetch a skill-convention README. With a slug, returns the per-app README. Without one, the org index.

clawnify docs                   # org-level: list available apps
clawnify docs my-app            # per-app README

clawnify mcp

Stdio MCP bridge to mcp.clawnify.com. Reads JSON-RPC frames from stdin, forwards each via HTTPS using your stored credentials, writes the response to stdout. Works with any MCP client that spawns subprocesses.

echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | clawnify mcp

Common wiring — drop into Claude Desktop / Cursor / Continue / Cline / opencode / OpenClaw config:

{
  "mcpServers": {
    "clawnify": { "command": "clawnify", "args": ["mcp"] }
  }
}

Full recipes in MCP gateway → CLI bridge.

clawnify sdk inspect <slug>

Developer-shaped alias for clawnify docs <slug> — same output, external density.

clawnify sdk inspect my-app

clawnify sdk login

Alias of clawnify login — re-authenticate or rotate the stored OAuth grant.

clawnify sdk login

clawnify --mcp

Starts the CLI as an MCP server exposing local deploy tools to AI coding agents. Different from clawnify mcp (the stdio bridge) — --mcp exposes deploy / list_apps / etc. locally; mcp proxies to the gateway. See the MCP page for wiring instructions.