Lookout — AI skill for Claude Code

👀 Lookout

Your screen-aware assistant. Watches your screen, answers questions about what it sees, alerts you when things change, and (with permission) clicks and types to do things for you.

Lookout is a local-first "computer use" agent you can actually read and modify. It pairs a vision model (Ollama by default — free & private — or Gemini / OpenAI / Claude in the cloud) with mouse + keyboard control, so you can have a conversation with your desktop.

+----------+        +---------+        +-------------------+
|  screen  |  --->  | capture |  --->  | vision model      |
+----------+        +---------+        | (Ollama/Gemini/   |
                                       |  GPT-4o/Claude)   |
                                       +---------+---------+
                                                 |
                              +------------------+------------------+
                              |                  |                  |
                         +----v-----+      +-----v-----+      +-----v-----+
                         |  ask     |      |  watch    |      |  act      |
                         | one Q    |      | monitor   |      | plan+do   |
                         +----+-----+      +-----+-----+      +-----+-----+
                              |                 |                  |
                              +--------+--------+--------+---------+
                                       |                 |
                                  +----v----+      +-----v-----+
                                  | answer  |      | pyautogui |
                                  +---------+      +-----------+

What you can do with it