Polar Cloud MCP
Official# Polar Cloud MCP
Stdio MCP server for [Polar Cloud](https://polar3d.com). Generate 3D models, slice with Polar Cloud defaults, and queue prints from Cursor, Claude Desktop, and other MCP clients.
Jobs are queued only. Start the print in Polar Cloud. This does not work in Claude.ai chat.
## Install
1. In Polar Cloud, open **Account Settings → AI Connect** and generate an API key.
2. Pick a client:
### Cursor
Click **Add to Cursor** on that same page. The key is filled in for you.
Or install this repo as a Cursor plugin, then paste the key under **Plugins → Configure**.
### Claude Desktop
Download [`polar-cloud.mcpb`](https://github.com/Polar3D/polar-cloud-mcp/releases/latest/download/polar-cloud.mcpb), open it, and paste the key when Claude asks. Claude stores it in the OS keychain.
### Other MCP clients
```json
{
"mcpServers": {
"polar-cloud": {
"command": "npx",
"args": ["-y", "github:Polar3D/polar-cloud-mcp"],
"env": {
"POLAR_API_KEY": "pc_live_…"
}
}
}
}
```
Optional: `POLAR_API_URL` (defaults to `https://polar3d.com`).
## Tools
Individual: `get_account`, `list_printers`, `generate_model`, `upload_model`, `slice_and_queue`, `get_job`.
Teacher / school admin: `list_school_queues`, `list_school_printers`, `preview_plate`, `approve_and_queue`, `reject_queue_item`.
## Local development
```bash
npm install
POLAR_API_KEY=pc_live_… POLAR_API_URL=http://localhost node index.js
```
Pack a Claude Desktop extension:
```bash
npx @anthropic-ai/mcpb pack
```
TDQS
Scored across 11 tools
Every tool has a clearly distinct purpose. Personal and school contexts are cleanly separated, and within each context, operations like listing, uploading, slicing, and approving are unambiguous.
All tool names start with a verb and use snake_case, which is consistent. Minor deviations like 'preview_plate' (verb+noun) and 'slice_and_queue' (compound verb) are still readable and follow the overall pattern.
11 tools is well-scoped for a 3D printing cloud server covering both personal and educational workflows. Each tool addresses a specific need without redundancy.
The tool set covers model generation/upload, slicing, and queuing, but lacks critical lifecycle operations: there is no way to list queued jobs, cancel a job, or start a print. Agents cannot fully execute a print workflow, leaving dead ends.