Your AI agents, in your pocket
Run Claude Code, OpenCode, or Codex on your dev machine — then watch the live terminal, send prompts, and approve or deny every action from your phone.
$ cadero start --agent claude
▐ claude · cadence
▸ wants to run: rm -rf ./dist
ApproveDeny— from your phone
Start a session
cadero start spawns your agent in its own room and prints a QR. Run one per project — the phone keeps them all connected.
Scan the QR with your phone
Open the PWA on your phone and scan the terminal QR with its built-in scanner. Keys are shared directly between your devices — the relay only ever sees ciphertext.
Stay in control
Live terminal, prompts, and an approve/deny dialog for every command the agent wants to run. Miss nothing on the go.
Security model
Zero-knowledge relay: a per-session AES-GCM-256 key is generated on your machine and handed to your phone exclusively via the terminal QR — the relay sees only room ids and ciphertext. The daemon never executes shell commands from the network: your phone sends high-level intents, the daemon validates them and feeds your agent's stdin.