polyshVideos keep pace with code
Make the walkthrough part of the same task, not a job that waits weeks for a free slot.
Let your coding agent make the product video.
Your coding agent already knows which screens changed. Ask it for a walkthrough, and it records the flow in your browser and sends Polysh every click and narration line. You get back a narrated video, captions, a guide, and a share link.
Early access. Nothing to install yet.
The agent that built it can show it
Polysh lets the agent that shipped a feature record its walkthrough while the context is fresh.
Small UI changes stop costing a re-shoot
When a button moves, the agent records the flow again and Polysh rebuilds the video and guide.
People still sign off
Every agent-made video lands in your workspace like any other, where your team can edit the script before sharing.
How your coding agent makes a product video
- 1
You ask
Name the flow, the audience, the languages, the URL, and the test account to use.
- 2
Your agent drives your browser
It clicks through your app on your own machine and records the screen.
- 3
It sends the bundle
The recording goes to Polysh with each click's time and position, the page URL, and narration per step.
- 4
Polysh returns the finished set
A narrated video with captions, a step-by-step guide, any translations, and a share link.
Your agent, your machine
Hosted recording agents run a browser on a remote server, which needs a public address and your logins. Polysh for agents works the other way around.
A hosted recording agent
- Browser: On a remote server
- Can open: Addresses the hosted service can reach
- Logins: Given to the service
- Click data: Depends on the service
Polysh for agents
- Browser: Yours, on your machine
- Can open: A local dev server, staging, or anything your browser opens
- Logins: Stay with you. The agent uses a test account you pick
- Click data: Logged by your agent at each click
Five ways to connect Polysh
Early access
Video agent
A skill for Claude Code, Codex, and Cursor that records your product and sends the result to Polysh.
Explore the Video agent
Early access
MCP server
Polysh tools your agent can call to upload, create, translate, export, and share videos.
Explore the MCP server
Early access
Skills
An open skill that teaches any skills-capable agent how to plan and narrate a good walkthrough.
Explore skills
Early access
CLI
Run Polysh from scripts and CI, for example to refresh your product videos on every release.
Explore the CLI
Early access
API
REST endpoints with API keys, webhooks for finished jobs, and an OpenAPI spec.
Explore the APIQuestions about agent access
Which agents can use Polysh?
Claude Code, Codex, and Cursor come first. Other agents that support skills can use the open skill, and one-click connectors for Claude and ChatGPT come later.
Does my code go to Polysh?
No. Polysh receives only the recording and its metadata: click times and positions, page URLs, and narration. What your agent shares with its own model provider is covered by your agreement with that provider.
What does it cost?
Polysh is free during early access, and agent jobs use the same workspace credits as the web app, charged only when a job succeeds.
Can I run it in CI?
That's what the CLI and API are for: starting Polysh jobs from a pipeline, such as on every release tag. Both are in early access.
Make the video part of the task
Join the agent waitlist and tell us which agent your team uses. Founding teams keep their rate.