Overview

ScreenCI makes product videos by prompting a coding agent. You describe the flow you want in a sentence, paste one prompt into Claude Code, Codex, Cursor or a similar agent, and it drives your real app, adds narration and camera motion, and hands back a polished video. When the UI changes, prompt it again (or rerun the recording) and the video regenerates.

How it works

Record and refine with the CLI and the web editor, then export and deliver the final video where your team needs it.

Capture

The ScreenCI CLI runs on your machine or in CI, capturing the product flow and uploading the raw recording plus timing data.

Render

ScreenCI renders that raw recording in the cloud, syncing narration, zooms, and animations to the timing data.

Only the raw recording and timing data leave your machine, never your source code, browser session, or credentials.

Deliver

Download the finished video and subtitles yourself, or publish them to a CDN-backed static URL that always serves the latest approved version.

Record and refine. screenci preview captures the flow on your machine, uploads the live preview, and opens the video in the Editor: a live preview and a timeline. Every change you make in the browser (narration, overlays, cuts, timings, render options) applies to the preview and later exports right away.
Export. When the preview looks right, run screenci export (or use the Export button) to render the finished videos and download them, then serve them from a stable public URL. export re-records only the videos whose sources changed, so it is also the command CI runs.

The CLI runs everything that touches your app locally, so what leaves your machine is the raw recording plus timing data, never your app’s source code. Projects created from the web app also upload their screenci/ scripts so the next editor can pull them (see Create Videos from the Web App); a repository-managed project keeps its scripts local unless it opts in. The CLI itself is open source.

Two ways to start

Both hand the work to your coding agent. Pick the one that matches where you are standing.

From the web app (no repository needed). Click Add project, paste the prompt into your coding agent, and the browser opens the video when the recording lands. Anyone in your org can add or change videos this way, including people who never open the code.
Agent integration. Already in a repository? Paste the integration prompt into the agent there. It scaffolds ScreenCI, authors a video for the flow you describe, and records it, with you mostly reviewing the result.

Tell the agent once what it needs to know about your product (repository, site URL, whether the site needs a sign-in, package manager, team notes) under AI context, and every later prompt starts from it. Set the look and voice new videos start from under Branding.

Once a video exists, refine it in the Editor or by prompting the agent again, and add Narration detail, Camera and Zooming, Overlays, or extra Languages.

Videos as E2E tests

Under the hood every video is a .screenci.ts file driven by Playwright, so you never depend on the agent to reproduce a recording. Ask the agent to commit the scripts into your repository and each video becomes an E2E test your own CI can rerun on every release: when the flow breaks, the run fails; when the UI changes, the video re-renders. That path (writing the scripts yourself, running them locally, and wiring up CI) is covered in Manual setup & first video, Video Script Basics and CI Setup.

Community

Need support, hit a bug, or want to share what you recorded? Join the ScreenCI community on Discord to get help, report issues, and talk to other users.