# CopperOS MCP server

> An open source agentic browsing assistant that reads, clicks, and types in Chromium-based browsers using accessibility trees and vision.

- Repository: https://github.com/Jesse0502/CopperOS
- Page: https://hostedmcp.org/servers/jesse0502/copperos
- Transport: —
- Stars: 10
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `AWS_PROFILE` | no | no |  |
| `STUB_FORM` | no | no |  |
| `STUB_NEW_TAB` | no | no |  |
| `STUB_NO_APPROVE` | no | no |  |
| `STUB_PASTE_REJECT` | no | no |  |
| `STUB_WEAK` | no | no |  |
| `TRANSCRIPTS_BUCKET` | no | no |  |
| `UNSUBSCRIBE_URL` | no | no |  |

## FAQ

### How does CopperOS interact with web pages?

It uses an accessibility-tree first approach for cheap and exact perception, escalating to screenshots automatically if the tree is unreliable.

### Which AI models can I use with CopperOS?

You can use CopperOS's hosted model, or run it locally/via API using Ollama, OpenRouter, or OpenAI.

### Does the agent use my actual browser session?

Yes, it reads, clicks, and types in your own tabs while you are signed in as yourself.
