Screenery
Publishes screenshots your tests already capture to stable @latest URLs so docs, READMEs and marketing pages update on merge, with optional visual regression checks.
At a Glance
Publish screenshots from CI to stable URLs. Full plan limits are unpublished.
Engagement
Available On
Alternatives
Listed Nov 2026
About Screenery
Screenery is a screenshot registry for product docs and marketing. CI pushes the screenshots that tests already take, and each one gets a stable URL ending in @latest.png whose image follows the default branch. The merge that changes a screen therefore updates the picture everywhere it is embedded, such as READMEs, docs sites and landing pages.
What It Is
Screenery works as a package registry for product screenshots. A CI step, the screenery/push@v1 GitHub Action or the screenery CLI, uploads PNGs (and GIF/MP4 clips) from a test output folder. Each asset is served from a CDN at a URL of the form cdn.screenery.dev/{org}/{project}/{asset}@latest.png. You paste the Markdown embed once, and the next push on the default branch changes the image behind it. Pull requests get their own @pr-N channel and do not move @latest.
Visual testing and review
Because a self-updating screenshot could publish a spinner or an error toast, Screenery offers an optional visual check. With a committed baseline set through diff-baseline, the Action compares every screenshot pixel by pixel before publishing, and a change past the threshold fails the job so nothing is pushed. Reviewers, or agents over MCP, can approve or reject each changed screenshot, and the result is reported to GitHub as a check named screenery that can be made required. Pull requests also receive a comment showing what changed against @latest.
Setup and credentials
On GitHub Actions the Action exchanges a GitHub OIDC token for a publish credential that lives five minutes, so no secret is stored. Other CI systems use a static publish token. Projects bind to a repository's numeric id, so renames do not break publishing and fork pull requests cannot publish. Assets can be public, unlisted (signed URL) or private (read token).
Agents and framework support
Screenery ships an MCP server that can create a project, push a first screenshot and wire the workflow from a single instruction in hosts such as Claude Code, Codex, Cursor and VS Code. Docs cover Playwright toHaveScreenshot goldens, Argos upload folders, Cypress, Vitest browser mode, Storybook, and Docusaurus and Starlight plugins. Further features include demo clips, LLM UI evals, live-page monitoring and Slack, Discord or Teams notifications.
Community Discussions
Be the first to start a conversation about Screenery
Share your experience with Screenery, ask questions, or help others learn from your insights.
Pricing
Free
Publish screenshots from CI to stable URLs. Full plan limits are unpublished.
- Screenshot URLs that follow your default branch
- Separate screenshot channels for pull requests
- Optional visual regression checks before publishing
Capabilities
Key Features
- Stable @latest screenshot URLs that follow the default branch
- GitHub Action (screenery/push@v1) and CLI for publishing from CI
- GitHub OIDC publishing with no stored secret
- Pixel-diff visual regression gate against a committed baseline
- Approve or reject changed screenshots, with a required screenery GitHub check
- Pull request comments with changed-screenshot previews
- @pr-N channels that do not move @latest
- Public, unlisted and private asset visibility
- MCP server for agent-driven setup and review
- Demo clips as GIF and MP4, LLM UI evals and live page monitoring
- Chat notifications for Slack, Discord and Microsoft Teams
- Docusaurus and Starlight docs plugins
