Privacy Policy

We use privacy-respecting, aggregate analytics (via Vercel) to understand which pages are read. It does not build advertising profiles and we don't sell or share it. The site stores two small preferences in your browser's local storage — your light/dark theme and, if you enter one in the demo banner, the name you'd attach to feedback. Both stay on your device.

When you deploy the Preview Dog banner on your own preview builds and a reviewer sends a note, the banner attaches the context they chose to share: the page URL, viewport size, browser and OS, the current branch and commit, and the note itself.

The server that receives the note also records the reviewer's IP address (read from the request) and stamps the issue with its source(“Preview Dog”), so the build team can tell where a report came from. Both are written into the same issue alongside the rest of the context.

That information is sent to your connected GitHub repository as an issue, using a token you control. It is your data in your systems — Preview Dog is the courier, not the destination.

On a deployment that turns on commenting and approvals, each time a reviewer signs in Preview Dog appends a single line to a private sign-in log — a dedicated tracking issue in yourGitHub repository, hidden from the banner's issue list. Each entry records the reviewer's name, email, IP address, browser, and the time, so the operator has a simple audit trail of who accessed the preview.

Like everything else the banner writes, this log lives in your repo under a token you control — you can read or clear it any time, and a deployment can leave the feature off. It never comes back to us.

The banner has an optional “Ask AI” chat that answers questions about the issues on a build. On this site the demo answers locally, with no third party involved. On a real deployment that connects an AI model, your question, any context you choose to add, and the list of issues are sent to an AI provider to write the answer. Preview Dog's built-in connector routes these through OpenRouter (an AI gateway); the deployment supplies the API key and picks which model answers. Ask is read-only: it never files, edits, or closes anything on its own.

When a reviewer signs in, Preview Dog keeps their recent Ask exchanges so the chat picks up where it left off after a reload or on their next visit. This is stored two ways, both as comments on dedicated tracking issues in your GitHub repository, keyed by email. The first is a conversation thread — the last handful of turns (roughly a dozen), kept verbatim: your questions and the AI's answers, word for word, capped to the most recent turns so it never grows without bound. The second is a short memory — on by default — a few plain-language lines summarizing what a reviewer tends to focus on, so later answers start warm rather than cold. To write the memory, the chat (your questions andthe AI's answers) is summarized by that same AI provider when the chat closes.

Both the verbatim thread and the summary live in your repo, keyed by email and read back into that person's future chats — you can read, edit, or delete either any time, and a deployment can turn the feature off entirely.

On a customer's own deployment, these conversations stay between the reviewer, the GitHub repository they're filed in, and the AI provider that deployment connected — they don't come back to us. Where we can see how Ask is used (like on this site), we may use it to make Preview Dog's prompts and defaults better. We never sell these conversations or use them for advertising.

No advertising trackers, no selling of personal data, and no third-party marketing cookies. The site sets no tracking cookies at all.

We rely on Vercel (hosting and aggregate analytics) and, for the banner integration, GitHub (where feedback issues and memory live). When a deployment turns on Ask AI, it also connects an AI provider — Preview Dog's built-in connector uses OpenRouter as the gateway — which processes the chat (the questions, the answers it writes, and the brief summary it distills for memory) under its own policy. The verbatim conversation thread and the summary are both stored in your GitHub repository. Each of these has its own privacy practices.

You can clear the site's stored preferences any time by clearing your browser storage. For anything else — questions, requests, or concerns — email hello@creo.design.

We'll update this page if our practices change and revise the effective date above. Continued use of the site means you accept the current version. See also our Terms of Service.