Legal
Privacy
Last updated: September 13, 2026
Signing in and Google user data
Signing in (email link, Google or GitHub) creates an account. When you sign in with Google, we receive your email address, name and basic profile information (such as your avatar) from your Google Account. We use this only to create and identify your Eigma account and to provide an automatic cloud backup of your wiki. This data is stored with our infrastructure provider (Supabase) and is readable only by your account. We do not sell it, use it for advertising, or share it with third parties. Eigma's use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.
Age
Eigma is not directed at children. You need to be at least 16 to have an account, the age Germany set for a minor's own consent to an online service under the GDPR. We do not knowingly collect personal data from anyone younger, and if we learn we have, we delete it.
Connecting an AI assistant
Eigma can be connected to an AI assistant such as ChatGPT or Claude through Eigma's MCP connector, so you can ask that assistant to file a conversation into your wiki or to propose an edit. You start the connection inside the assistant and authorise it with your Eigma account over OAuth. The authorisation is per assistant, and you can withdraw it at any time in Eigma under Settings, Connected assistants. If you never connect an assistant, none of the connector behaviour described here applies to you.
What a connected assistant sends us
When you ask a connected assistant to save something, it sends Eigma the pages it drafted for you. For each page that is a title, the chapter it should live under, a short summary, the body text in Markdown, optional tags and links, and the quiz questions it wrote for that page. It also sends a small provenance block about the source chat: its title, its date, and a one line summary. We do not ask for the chat transcript and we do not keep it. Our server keeps only those three provenance fields and discards everything else that arrives in that block before the draft is stored, so your raw conversation never lands in our database.
What the connector returns to the assistant
So that an assistant can update the right page instead of duplicating it, Eigma's read tools hand your own wiki content back to it. Browsing returns your chapter names and page titles. Searching returns matching page titles with their chapter and summary. Reading a page returns that page's identifier, title, chapter, summary, full body text, tags, how many quizzes it has, and each quiz's identifier and question text, plus the title of the conversation the page came from; quiz answers and answer options are never returned. The review list returns the name and language of your wiki, counts and progress figures derived from it, your chapter names, and the identifier, title and chapter of the pages that are due to read, to test, or fading. Saving pages or proposing an edit returns a receipt naming the pages involved. The connector reads all of this from your cloud backup, so it can only see what you have signed in and synced.
Your assistant provider sees what the connector returns
Everything described above travels to the assistant you connected, which means OpenAI for ChatGPT and Anthropic for Claude, and is then handled under that company's own privacy policy and your settings there. This is inherent to how a connector works: the assistant has to see a page in order to help you with it. If you would rather it did not, do not connect an assistant, or disconnect the one you have.
Nothing enters your wiki without your review
A connected assistant cannot write to your wiki. Every save and every proposed change is queued in Eigma's review inbox, where you see it next to the current page and accept or reject it yourself. Queued items you never accept stay in the inbox until you delete them. A draft of a proposed change that you never accept is discarded from our server once it is more than seven days old.
Publishing to the community
Your wiki is private, and nothing in it is public unless you publish it. If you do publish a volume, a read-only copy of its pages and quizzes goes on eigma.app where anyone can read it, under the name you chose to publish with. Ratings, reviews and comments you leave are public in the same way and carry that name; a report you send us is not, and only we see it. Unpublishing a volume or deleting a rating removes it from the site, and deleting your account removes all of it.
Connector diagnostics
For each request an assistant makes to the connector we store one diagnostic row: the endpoint that served it, the assistant's user agent string, the session identifier its client sent, the name of the tool it called, the HTTP status, the duration, the response size, and, when the call failed, up to 300 characters of the error message that tool returned. That error text is written for people to read, so it can name a page you asked about. These rows hold no page bodies and are readable only by us, and we use them only to diagnose connector faults. We keep these rows for no longer than 30 days, after which they are deleted automatically.
Where the connector runs
The connector is part of Eigma's own API, hosted with Vercel, and everything it stores lives with Supabase, the same infrastructure provider that holds your cloud backup. Those two are our only processors for connector data. Supabase also sends the sign-in emails, through Resend. Your assistant's provider is not our processor: you connect Eigma to it yourself, and what it does with what the connector returns is governed by its own policy, not ours. Both are in the EU, and in fact both in Paris: the connector runs as a serverless function in Vercel's Paris region and keeps nothing there, and everything it stores goes to the Supabase project, which is hosted in Paris as well.
How your data is protected
Every connection to Eigma, from your browser, the desktop app or a connected assistant, runs over an encrypted connection (TLS). Once you are signed in, your cloud backup and anything you publish are stored with Supabase, which encrypts data at rest as part of its own infrastructure; we do not add our own encryption on top, and Eigma's own systems can read it, so this is not end-to-end encryption. The copy of your wiki in your browser or on your device is not separately encrypted by Eigma; its protection depends on your own device's disk encryption.
Deleting your data
You can delete your account and all associated cloud data at any time, yourself: open Settings, then Account, and use the delete option at the bottom of that screen. It asks you to type a confirmation word first, because it cannot be undone. If you would rather we did it, email support@eigma.app. Deletion removes your backup from our infrastructure, together with anything a connected assistant has queued for your review, any pending change drafts, and that assistant's authorisation to reach your account. The connector diagnostic rows described above carry no account link, so deleting your account does not reach them; they are deleted automatically within 30 days of being written.
How long we keep things
Your knowledge base has no expiry: we never delete it on our own, and it is yours to keep or remove. If you sign in, one backup copy of it is stored for your account and is replaced each time it syncs, so there is one current copy and no history of older ones. Items an assistant has queued for your review stay until you accept or discard them, or until you delete your account. Connector diagnostics are kept for no longer than 30 days. Deleting your account removes the rest, as described below.
Why we are allowed to process this
For people in the EU and UK, we rely on two grounds. Providing the service you asked for covers nearly everything here: your account exists so you can sign in, your backup exists because you turned it on, and the review inbox exists because you connected an assistant. Keeping the service secure and working covers the rest, which is the connector diagnostics. We do not rely on consent for any of it, because none of it is optional extra processing, and we do not process any special category of data.
Your rights over your data
What we hold about you is your account details: the email address you signed in with, your name and avatar if your sign-in provider supplied them, and the backup copy of a knowledge base you already have. You can ask us to correct any of it or to delete it. Mostly you do not have to ask: your knowledge base exports from the app in one click, in a portable format you keep, and account deletion is a button rather than a request. If you would rather write to a person, use the address below. If you are in the EU or UK and you think we have handled your data wrongly, you can complain to your national data protection authority.
Who is responsible, and how to reach us
Eigma is built and operated by Ya Wang, an independent developer in Germany, who is the controller of the data described here. For a correction, a deletion, or a question about this policy, write to support@eigma.app.
Measuring how the site and the app perform
This website loads Vercel Web Analytics, and the web version of the app loads Vercel Web Analytics and Vercel Speed Insights, which count page views and measure loading speed. They record the page you are on, not who you are, and set no cookies. Everything on this website is public, so the address of the page you are reading is what gets recorded. Inside the app it works differently: before a page view is sent, the address is rewritten so that a page of your own is reported as /node/:id rather than by its identifier, which means the names and identifiers of your pages never leave your device. None of it is used for advertising, and none of it follows you to other sites. All of it runs in the browser; the desktop app sends nothing.
Cookies
Eigma sets no cookies of its own, here or in the app. Staying signed in works by keeping your session in your browser's own storage, which clearing your browser data removes. There are no analytics or advertising cookies.
What we do not do
No ads, no selling your data, and no advertising or cross-site tracking of any kind. We do not train any model on your knowledge base, and we never will.
This policy will grow
Eigma is young. As features like Pro arrive, this page will describe exactly what changes, in plain language.