Privacy

The Atlas has no accounts, no sign-up and no forms. It does not ask for or store personal data. This page lists what the services it runs on can still see.

Last updated: 4 November 2026.

Who runs this

Oliver's mTOR Atlas is an independent, non-commercial project curated by Oliver Barton, Prague. The website, the JSON API, the MCP server and the ChatGPT plugin are published and operated by Petr Barton, Prague, on Oliver's behalf, because Oliver is not yet of legal age. Petr Barton is responsible for the processing described on this page. Contact: oliver.barton1113(at)gmail.com.

The website

• Hosting. The site is static and hosted on GitHub Pages. GitHub receives the usual request data (such as IP address and browser) to deliver pages and keep the service secure, under the GitHub Privacy Statement. The Atlas has no access to these logs.
• Usage statistics. Pages load Google Analytics 4 (property G-420TPC8J46) to count visits, pages viewed and where visitors come from. Each visit also reports the browser's User-Agent string and screen size, which the Atlas uses only to tell automated traffic from readers. Google Analytics sets first-party cookies and does not log or store IP addresses. Event data is kept for 2 months. The Atlas uses the reports only to see which pages are read; it does not use advertising features or combine the data with anything else. See Google's Privacy Policy; you can block it with any content blocker or the Google Analytics opt-out add-on.
• "Was this page useful?" If you answer, Yes or No is sent to Google Analytics as an anonymous event together with the type of page (for example "study"). Nothing else is asked, no text is collected, and your answer is remembered in your own browser only so the question is not repeated.
• Bluesky buttons. "Follow" and "Share" open bsky.app, which is run by Bluesky under its own privacy policy. The Atlas counts only that the button was clicked, not who clicked it.
• Practice and progress in the Academy are stored only in your own browser (local storage). They are never sent anywhere.

The JSON API

The API is a set of static files on the same GitHub Pages hosting. It needs no key and sets no cookies. Requests are seen only by GitHub, as above.

The MCP server

• Local version (npm package mtor-atlas-mcp). It runs on your own computer and sends nothing to the Atlas. It downloads the public API files from mtor-atlas.org, which GitHub sees as ordinary web requests.
• Remote version (https://mtor-atlas-mcp.mtor-atlas.workers.dev/mcp). It runs on Cloudflare Workers. When an AI assistant uses it, the server receives only the tool call: the tool name and its arguments, such as a search term or a study ID. It does not receive your conversation, your name or your account. It needs no sign-in and sets no cookies.
• What you ask is kept briefly and privately. For each tool call the server stores the arguments (for example the search term or study ID, at most 200 characters, with e-mail-like text removed) so the Atlas can see what people look for and improve its coverage. It does not record the results, your conversation, IP addresses or browser details, and the arguments are not published anywhere; only the Atlas maintainers can read them. Please do not put personal or health details of yourself or others into a search. The server stores nothing else between requests other than a temporary in-memory copy of the public Atlas data. Request logging in Cloudflare is switched off.
• Anonymous usage counts. For each request the server adds one count to an aggregate table (Cloudflare Workers Analytics Engine): the time, which tool was used (for example search_studies), the name the connecting app reports for itself (for example claude-ai), the kind of app taken from its User-Agent (for example "claude" or "openai"), whether the call succeeded and how long it took. Nothing in these counts identifies a person. The Atlas uses these counts only to see whether and how the server is used. Cloudflare keeps them for 3 months.
• Counting distinct users. Some apps send the server an anonymous ID with each tool call; ChatGPT and Codex do this (the openai/subject field), so a service can tell repeat users apart without knowing who they are. The server does not store this ID. It stores only a salted one-way hash of it, which cannot be turned back into the ID, so the Atlas can count how many different users the plugin had on a day, and nothing more. It also records the names (not the contents) of the technical fields an app sends. These are kept for 3 months like the other counts.
• Cloudflare itself processes network traffic (including IP addresses) to deliver and protect the service, under the Cloudflare Privacy Policy.
• What it returns is public metadata about published research (titles, authors, DOIs, curated findings), the same as on the website. It contains no personal or health data about users.

Sharing and selling

Nothing is sold, shared with advertisers or used to profile anyone. The only third parties involved are the service providers named above (GitHub, Google Analytics, Cloudflare, and npm for the package download).

Retention

The Atlas keeps no personal data of its own. Google Analytics event data is deleted after 2 months; the MCP server's anonymous usage counts and stored tool arguments after 3 months. Retention of operational logs at GitHub and Cloudflare is governed by their policies linked above.

Questions and changes

Write to oliver.barton1113(at)gmail.com. If anything here changes, this page is updated and the date at the top changes with it.