Privacy Policy

Last updated September 26, 2026

Ravensight is a game analytics service operated by Reality Software Entertainment, LLC, an Ohio limited liability company in Mount Vernon, Ohio, doing business as Ravensight ("we", "us"). This policy explains what data Ravensight handles, in two distinct roles: data about you, the developer using the dashboard, and data about players of games that integrate a Ravensight SDK.

Data about developers

Signing in uses RealitySE, our company identity service. Ravensight receives your account identifier and, where available, your email address; passwords are never seen or stored by Ravensight. We keep operational records tied to your account: the games you create, usage counts for billing, an audit log of dashboard actions, and support correspondence. Payment details are handled by Stripe; Ravensight never stores card numbers.

We use a third party product analytics service to understand how the dashboard is used. It records feature usage tied to your account identifier and plan. It never records your email, name, keys, tokens, or the content of analyst questions.

What the AI features send, and where

When a player explicitly submits a screenshot report, we store the submitted game image and the accompanying message, build, session reference and developer-selected context. Players can redact or omit the image before sending. Report images are private to authorized studio members and expire after 90 days; messages and issue history remain until removed. Screenshots are not automatically sent to AI providers.

Several Ravensight features work by sending your content to Anthropic, the provider of the Claude models: the questions you ask the analyst along with the query results it reads to answer them, design studio conversations and the documents they reference, player feedback messages when processed in an enrichment batch, and, if you run a source scan, the source files you submit. Player feedback text is also sent to an embedding provider so that feedback search can match on meaning rather than exact words. Selected evidence excerpts may also be processed by TypeSafe AI to classify their topic and relevance within Studio Intelligence. These providers process that content to return a result to us. Provider requests use minimized content rather than account credentials. Avoid including secrets or personal data in submitted content.

Source scans run when you request them and quote the price first. Background digests, pattern synthesis and feedback enrichment can also process relevant content without a new chat request; they are bounded by backend work limits.

Source scanning

A source scan reads the code you choose to give it, and then discards it. That code reaches us one of two ways: you upload it, or, if you have connected a repository, we read it from GitHub with the read-only access that installation grants us (file contents and metadata, nothing else, and we never write to your repository). You control that access from GitHub or by disconnecting in the dashboard, and revoking it stops us reading anything further. What we keep is our own notes: for each file, its path, a SHA-256 hash of its text, and a short description of what that file does and what telemetry it reports, plus a repository-level summary. We keep our notes, not your code. There is nowhere in our schema to store file contents, no diff and no patch, and a hash cannot be turned back into the file it came from. The hashes exist for one reason: so a rescan can tell which files changed without us having to keep the bytes, which is what makes a rescan cost a dollar instead of a full scan price.

Those notes have no expiry, deliberately, because losing them would silently re-charge you full price for work you already paid for. They are deleted when you delete the game they belong to, or when you disconnect the repository. Findings a scan produced are stored like any other finding in your studio context and are deleted with the game.

Ravensight Playtest

A playtest runs on your own machine. Your source code and your build are read locally and are never uploaded: no upload path accepts them, and the model proxy never receives them. Code reaches us only through a source scan you start yourself, which is the separate feature described above.

What does transit Ravensight, on its way to Anthropic, is what a persona needs in order to play: your brief and the persona instructions, the screenshots and state it observes turn by turn, and the report text it writes back. Every model call goes through our proxy, which rebuilds each request from a fixed field allowlist rather than forwarding yours. We do not store that content. What we keep is one metering row per call: which job and run it belonged to, which step, the model chosen, token counts, cost, latency, why the model stopped, whether it was served or refused, and which account pays. Those rows hold no prompt or completion text and are kept for 13 months, because the daily reconciliation and the provider invoice are both read against them.

What uploads to Ravensight as files is exactly what a fixed allowlist names: the reports, their structured findings, the screenshots they cite, and run bookkeeping. Session video and the full action transcript are opt-in and are not sent unless you ask for them. A file outside the allowlist is refused before a link is issued, and a file that trips our secret scan is quarantined rather than stored alongside your reports. Findings cite a file and a line range; the source text behind that line is not uploaded. The dashboard's What leaves your machine page lists the allowlist and the proxy rules against the code that enforces them.

A persona plays through your own Ravensight SDK, so its session writes ordinary gameplay events into your own event stream, tagged as synthetic. Those rows are yours and are subject to the same retention as any other event.

The Playtest CLI signs in with a device-code token scoped to the games you approve. It never holds a provider key. You can revoke a token from the dashboard, and an organization owner can revoke every token a member holds in that organization at once.

Reports, findings, marks and briefs are kept for as long as the game exists. Session video and the action transcript, when you opt in to them, are stored with a 90-day retention tag that the storage bucket's own lifecycle rule acts on, so their expiry does not depend on any of our jobs running. Deleting a game deletes its playtest records and its stored files with it, including the job records themselves. Two things survive deletion on purpose, and neither holds prompt or completion text: the month's billing history, and the per-call metering rows described above.

Data about players

Games that integrate a Ravensight SDK send gameplay events: an event name, a timestamp, an anonymous device identifier generated by the SDK, a session identifier, platform, game version, and whatever event data the developer chose to attach. Ravensight does not collect player names, emails, or contact information, and the SDKs provide no mechanism for doing so.

For player data, the developer of the game is the data controller and Ravensight is a processor: what ends up in an event payload is entirely the developer's choice. Developers who put personal data in event payloads are responsible for disclosing that to their players and for having a lawful basis to collect it.

Children

Ravensight is a service for game developers and is not directed to children; developer accounts are for adults. The SDKs collect no names, emails or contact details from any player, including young players. A developer whose game is directed to children is responsible for meeting children's privacy law for that game, including keeping personal information about children out of event payloads.

Retention

Raw gameplay events are kept for 90 days and then deleted automatically. Compact aggregates derived from them (daily and monthly totals such as active players, sessions, and event counts) are kept indefinitely so long-term charts keep working; these aggregates contain counts, not individual event records. Developers can turn history collection off per game at any time. Audit log entries expire after 180 days. Playtest session video and action transcripts, both opt-in, expire after 90 days; playtest reports, findings and briefs last as long as the game. Billing usage records, and the per-call playtest metering rows described above, are kept as long as the account exists and for 13 months respectively.

Sharing

We do not sell data, and we do not share it with third parties except the service providers that run Ravensight: cloud infrastructure (Amazon Web Services), event storage (ClickHouse Cloud), operational databases (MongoDB Atlas), payments (Stripe), email delivery (Amazon SES), AI models (Anthropic, TypeSafe AI for evidence classification, and an embedding provider for feedback search, as described above), and a product analytics service. Playtest persona prompts and observations reach Anthropic through our own proxy, which meters the call and keeps no content. Each processes data only to provide its service to us.

Where data is processed

Ravensight runs in the United States. If you or your players are elsewhere, data is transferred to and processed in the United States.

Security

Data travels over encrypted connections and is stored with providers that encrypt it at rest. Access is limited to the service itself and to the people who operate it. No system is perfectly secure; if a breach affects your data, we will notify you as the law requires.

Cookies

The dashboard uses a first-party session cookie to keep you signed in. The product analytics service described above may store an identifier in your browser. Ravensight does not use advertising cookies.

Your choices

You can delete a game at any time, which deletes its data; you can pause tracking or history per game; and you can contact support@realityse.com to ask about, correct, or delete data tied to your account. Depending on where you live, you may also have the right to receive a copy of your data or to object to how it is processed. We answer those requests at the same address, and making one will not change how we treat you. Players should direct requests to the developer of the game they played, who can act on the data they control.

Changes and contact

If this policy changes materially we will note it here with a new date. Questions go to support@realityse.com, or to Reality Software Entertainment, LLC, Mount Vernon, Ohio.