QuanterLab Desktop

The Primitives engine, running on your own machine. Long research walks use your CPU, not ours. Sign-in, market data and publishing stay with the platform.

Download for Windows
Version 0.1.5 · 286 MB · Windows 10 or 11, 64-bit · about 1.2 GB on disk once installed
This build is not yet code-signed, so Windows SmartScreen shows a warning on first launch. Choose More info, then Run anyway. A signed build follows and installs over this one through the app's own updater.

macOS: an Apple Silicon build is in preparation. It ships once it is notarized with Apple, and will appear here and in the app's updater.

You need a QuanterLab account with Primitives access to sign in. The engine will not fetch data without one. Follow the research to hear when access opens.

What runs where

The desktop is the third of three ways an AI can work with QuanterLab, and each does a different job. Expecting one to do another's job is the mistake this table exists to prevent.

Quantin, in the platformThe QuanterLab connector, in claude.aiQuanterLab Desktop, on your machine
What it isA chat assistant inside the web app.A hosted MCP server your claude.ai account connects to.The engine, a canvas, and a local MCP server your own AI connects to.
It canExplain primitives, methods and results; guide you around the platform.Read the catalog, list your own saved work, poll progress, summarise results.Build and run circuits, walk a registered study window by window, compile it, submit it to the platform, and load custom primitives you write.
It cannotRun anything. It is a guide.Start a run. Compute does not happen on our servers through it.Publish. The platform's review decides that. It also returns no price series.
Where the compute isNone.None.Your machine.
How to startOpen it inside the platform.Add the connector in claude.ai settings. Setup.Install, sign in, then add the MCP server to Claude Desktop or Claude Code (below).

Connect your AI to the desktop

After installing and signing in once, add the desktop's MCP server to the assistant that runs on your machine. The command is the bundled sidecar; the path below is the default install location.

Claude Desktop (settings, Developer, edit config):

{
  "mcpServers": {
    "quanterlab": {
      "command": "C:\\Users\\<you>\\AppData\\Local\\Programs\\QuanterLab Desktop\\resources\\sidecar\\quanterlab-sidecar.exe",
      "args": ["mcp"]
    }
  }
}

Claude Code:

claude mcp add quanterlab -- "C:\Users\<you>\AppData\Local\Programs\QuanterLab Desktop\resources\sidecar\quanterlab-sidecar.exe" mcp

The desktop app must be running: the server talks to the engine on your machine and refuses to serve until you have signed in. On macOS the sidecar is /Applications/QuanterLab Desktop.app/Contents/Resources/sidecar/quanterlab-sidecar and Claude Desktop's config file is ~/Library/Application Support/Claude/claude_desktop_config.json.

What the platform still does

Three things never leave quanterlab.com. Your sign-in, which issues the desktop a device token you can revoke. Market data, which the engine fetches through a metered proxy under that token, so no data key is ever stored on your machine. And publishing: a study you register, walk and compile on the desktop uploads to the platform as a draft with its full evidence, and the review gate decides what goes public.

Custom primitives

Drop a Python file into ~/QuanterLab/plugins, or let your AI write one through the MCP server. It registers through the engine's own decorator, has to use a category and port names the canvas already knows, and appears in the palette on the next reload. Your code, your machine; the engine itself ships compiled.