Chrome Extension, Google Meet

Record the meeting.
Send no bot.

Decker captures the Meet tab from your own browser. No extra participant joins, nothing for IT to block. While you talk, it builds the deck, brief, or prototype, ready to share before the call ends.

Get early accessOpen Source on GitHub

Not on the Chrome Web Store yet. Email for the build, or load it from source in two minutes.

Open source (MIT). Works with your own API keys. No account, no server.

What you get

Decker turns the conversation into something you can send: a working prototype of the feature you just spec'd, or a deck for the decision you just made, finished while you're still on the call.

How it works

The artifact builds while you talk.

By the time you hit Generate, the research is done.

Hit Record
One click in Chrome while you're live on Meet.
0:00
0:16
Topics appear
GPT-4o mini extracts discussion points from the live transcript.
Select a topic, research runs
Research starts in the background the moment you check it.
0:32
ongoing
Meeting continues
Topics refresh. Research completes. Nothing interrupts the call.
Hit Generate
GPT-4o builds the artifact. The research is already done.
end
+90s
Share before you hang up
Drop the file in Slack while everyone is still on the call.

Outputs

One meeting. Four kinds of artifact.

Every output is a self-contained HTML file. Open it, share it, drop it in Slack.

Prototype
A working product, not a mockup.

Clickable, interactive HTML built from what you described on the call.

Product owners, founders
Presentation
A deck without the slide software.

Self-contained HTML, keyboard navigable, share as a single file.

Engineering leads, PMs
Discussion SPA
A website about what you just decided.

Hero, sections, decisions, insights, a brief your whole org can read.

Stakeholder reviews
Meeting Brief
Decisions and actions, structured.

Per-topic sections, key decisions highlighted, full action item table.

Any technical meeting

Under the hood

Ready the moment you generate.

Every part of the pipeline runs in parallel so there is no wait at the end.

Live transcription

Whisper processes audio every 16 seconds. The transcript builds as you talk.

Parallel research

Select a topic and GPT-4o mini researches it immediately, while the meeting runs.

Four artifact formats

Prototype, deck, SPA, or brief. One self-contained HTML file, every time.

Tab and mic capture

Captures the Meet tab audio and your mic. Mixed and transcribed together.

State persists

Close and reopen the popup anytime. Topics and research are still there.

Runs entirely in Chrome

No backend, no servers. Keys stored locally. Nothing leaves your browser.

Who it's for

For the meetings where something real gets built.

The spec review. The architecture call. The session where you walk an engineer through a product idea for the first time. The meeting where you want to leave with something in your hands.

Product Owners
Engineering Leads
CTOs and VPs Eng
Founders
Design Leads
Technical PMs

Your own API key

One OpenAI key. It stays in your browser.

Calls go straight to OpenAI. We never see your audio, transcript, or key. There is no server to send them to. A full 60-minute meeting costs a few cents.

OpenAI Whisper
GPT-4o mini
GPT-4o

Works in

Chrome
Brave
Arc
Edge
Firefoxsoon
Safarisoon
Decker

Build the thing
before you hang up.

Free and open source. Bring your own API keys; a full meeting costs a few cents.

Get early accessView on GitHub

No subscription. Firefox support is in progress.