What is cursy?
cursy lets you control Cursor from iMessage. You text your own number (or a dedicated thread) and a small daemon on your Mac runs the prompt through the cursor-agent CLI, then texts the result back — like having a coding agent you chat with from your phone.
How does it work?
A background daemon (a launchd agent) watches the local Messages database for new texts from whitelisted handles. Each message runs through cursor-agent in your chosen workspace, and the reply is sent back through Messages.app. Every thread maps to a persistent agent session, so follow-ups keep full context.
What can I do with it?
Ask it to read, edit, and run code in a workspace repo — "list the largest files," "add a test for the parser," "fix the failing build." It's a coding agent scoped to your repo, not a general chatbot. Built-in commands like /new, /workspace, and /model let you steer each thread.
Is it private?
Yes. Everything runs locally on your Mac — no cloud relay, no third-party servers — and your iMessage credentials never leave Messages.app. cursy reads the Messages database read-only and replies via AppleScript.
Is it safe?
Only whitelisted handles are processed, and SMS/RCS is ignored by default (iMessage only). You can require a passphrase, scope the agent to a dedicated workspace, or disable write/shell access. With the default force mode it can edit files and run commands, so treat it like SSH access to your machine.
What do I need?
macOS 14 (Sonoma) or newer, Messages signed in to iMessage, and a Cursor account (cursy can install the cursor-agent CLI for you). Setup walks you through two macOS permissions. Install with the command above or brew install pontusab/tap/cursy, then run cursy start.