Introduction

Lumo is a browser-native toolset. It does not replace the browser or move your work into another app. It adds two precise capabilities to the browser you already use: understanding and control.

Two products, one boundary#

The products are separate on purpose. Each has its own permissions, storage and release cadence. They share a design language and a local-first stance, not a runtime.

• Lumo — an AI sidebar with a full agent loop: it reads the current page, calls browser and DevTools tools through MCP, and keeps working until the task is done.
• Lumo Userscripts — a userscript manager built on native chrome.userScripts, with a real editor, GM APIs, backups and diagnostics — plus an MCP endpoint the agent can operate.
• This website — the brand surface, product documentation and the reference implementation of the shared design system.

What makes it different#

• Local-first. Conversations, files, settings and script bodies stay on your device. There is no Lumo account and no relay server.
• Bring your own model. Lumo talks to the provider you configure, using your key. A local model is a first-class option.
• Native platform APIs. The userscript manager registers scripts through chrome.userScripts; it does not simulate them with content-script injection.
• Agent-native, not agent-wrapped. The Lumo Userscripts endpoint is standard MCP, so the agent creates and edits scripts instead of instructing the user.
• Craft as a constraint. Four design laws — invisible assistance, coherent familiarity, progressive disclosure and cinematic momentum — are part of the review checklist.

Where to go next#

1. 1

   Install a product

   Start with Lumo for page understanding, or Lumo Userscripts for per-site control.

2. 2

   Read the architecture

   See how the extension runtime, the model layer and this static site fit together.

3. 3

   Use the design system

   Tokens, palettes, motion and accessibility rules are documented as executable constraints.