How Lumo works

Lumo records a narrated walkthrough of your screen. You talk while you point at things; when you press stop, a link is on your clipboard about one second later. Paste it into any AI chat — Claude, ChatGPT, Cursor, or any agent that can fetch a URL — and the AI gets a structured, timestamped account of what you saw, what's wrong, and what "done" looks like, with screenshots.

No video uploads, no manual screenshot dragging, no retyping the problem. One shortcut, one link.

What the AI receives

Every share link is readable without JavaScript, and exposes three machine-friendly views:

/share/{slug} — the human-readable walkthrough page.
…/prompt.md — an agent-ready markdown prompt: the problems, environment, and desired outcome.
…/manifest.json — the structured package: timestamped transcript segments, screenshot references, pointed-at moments.
…/bundle.zip — everything above plus the screenshots, as one archive.

Shortcuts

⌃ ⌃ (double-tap left Control)	Start a take — the default gesture; tap ⌃ once to stop
⌃⌥⌘L	Start / stop — the chord alternative
⌃⌥⌘X	Cancel the take — nothing is published
⌃⌥⌘C	Copy the last link again
⌃⌥⌘Space	Open the Lumo panel
⌃⌥⌘R	Restart the take

Every shortcut is remappable in Settings.

Good to know

Requirements: macOS 14+, Apple silicon. The app is signed and notarized by Apple.
Speed: the link is on your clipboard about one second after you stop (p50 ≈ 0.94s).
Privacy: transcription runs on your Mac. A link publishes only what you narrated over: screenshots taken during the take plus your words. Links expire within 24 hours unless saved to an account.
Accounts: sign in at lumo.link and link your Mac (Settings → Account) to keep a dashboard history of your links.

Get Lumo

Lumo is in private early access on macOS.

Download for macOS