Skip to article
Docs / CLI

CLI

Manage your brands, approve posts, and chat with AI — all from the terminal.

Run from source

These local examples require a configured app checkout, app and CLI dependencies, and Bun. Start the app with npm run dev from the repository root and wait for its ready message. Use its actual origin: http://localhost:5173 below is only an example, not an assumed running service. Run CLI commands from cli/. Set PUBLIC_APP_URL before login and every command, and set the same value in each separately launched MCP process's env configuration. Without this pin, local detection can fall back to production. The MCP listener on port 8787 is not the app origin. No hosted installer or published package is assumed.

export PUBLIC_APP_URL="http://localhost:5173"
bun run cli.ts --help
bun run cli.ts brands

The local package declares crumbless and crumbless-mcp. The examples below use those names when available on PATH. Otherwise replace crumbless with bun run cli.ts from cli/.

Authenticate

Run crumbless login to sign in through the browser. Local MCP shares this saved session. Set PUBLIC_APP_URL to your intended app origin before logging in or running commands, especially when switching between development and production.

export PUBLIC_APP_URL="http://localhost:5173"
bun run cli.ts login

Quick commands

export PUBLIC_APP_URL="http://localhost:5173"
crumbless brands
crumbless dashboard <slug>
crumbless content <slug> --status pending_user
crumbless approve <slug> --all
crumbless post <slug> <id> edit --caption "..."
crumbless plan <slug>
crumbless ai <slug> --message "..."

Updating

For a source checkout, use the version supplied by your repository maintainer. Check the command's help against that checkout before using it.

AI Agent skill

The repository includes instructions for coding agents. Configure the skill and MCP or CLI access using the agent integration guide, then ask for a specific task:

  • "Use the Crumbless CLI to show me my brands"
  • "Approve all pending posts"
  • "Ask Crumbless to analyze my posts"

Next

  • MCP — connect Cursor or other agents via Model Context Protocol
  • Agent skill — teach agents when to use MCP or the CLI
  • See the API reference for building custom integrations.