Atlas logoAtlas

Privacy Policy

Last updated: May 25, 2026

Atlas (“we”, “us”) is operated by VexHex Games, Inc. We are a small team building an open Catan and Landover game engine. This policy explains what data we collect and how we use it.

TL;DR

  • We don’t set tracking cookies and don’t run analytics.
  • We don’t collect personal information from people who just visit the site.
  • If you run an Atlas worker, we receive the games it plays plus the node-ID label you chose for it. Nothing else.
  • We never sell data. Game data is used to train the Atlas model.

1. What we collect

When you visit the site

Our hosting providers (Vercel for the website, Cloudflare for the CDN, Fly.io for the API) record standard request logs containing your IP address, browser user-agent, and the URLs you access. These logs exist for security and operational reasons. We do not access or analyze them individually.

When you run an Atlas worker

Your worker uploads the games it played: seed, agent configuration, action history, who won, and the node-ID label you chose inatlas-worker.json. The shared contributor API key (atlas-contributor-public) is not tied to your identity. We do not collect your IP address against your node-ID, your email, or any other personal identifier.

The node-ID is displayed publicly onatlas.landover.io/workersalongside the count of games that node contributed. If you used a default like your computer’s hostname, that hostname will appear publicly. Set thenodeId field to anything you want.

2. Cookies

We do not set tracking cookies, advertising cookies, or analytics cookies. Cloudflare and Vercel may set strictly necessary cookies for bot-protection (__cf_bm) and session routing (__vcrcs); these are exempt from consent requirements under GDPR and CCPA because they are required for the service to function.

3. How game data is used

Uploaded game trajectories are used to train Atlas’s neural network and improve the engine. Aggregated counts (total games, positions, workers) are shown publicly on the homepage. Individual trajectories are not made public, but trained model files are.

4. Data retention

Worker trajectories are stored in Cloudflare R2 indefinitely as part of the training dataset. We may delete or compact older trajectories as storage or model needs change.

5. Third-party services

  • Vercel — website hosting (US).
  • Cloudflare — CDN, DNS, and R2 object storage (US/global).
  • Fly.io — API hosting (US).
  • Upstash — Redis for counters (US).

Each of these providers maintains its own privacy policy governing the minimal data they handle on our behalf.

6. Your rights

Because we do not knowingly collect personal data, there is typically nothing for us to delete or correct on a per-person basis. If your chosen nodeId contains personal information you regret publishing, contact us and we will remove or rename it.

7. Changes

We will update this page when relevant practices change. The last-updated date at the top reflects the most recent revision.

8. Contact

Email legal@landover.io for privacy questions, data requests, or to report a concern.