NODUS · 0.3.0

Privacy and liability.

Short version: your records are in your git repository, the engine keeps no copy of them once a request ends, and nothing here is guaranteed against loss. The long version is below.

No warranty, and no liability for lost records

Nodus is provided as is, without warranty of any kind. It is young software and it can have bugs.

We are not liable for any loss of your records — not for data that goes missing, is overwritten, is deleted, fails to sync, or cannot be read back, and not for anything that follows from that. This holds whether the cause is a defect in Nodus, a mistake by an AI writing into your notebook, your git host, your own machine, or anything else.

Two things follow from how Nodus is built, and both are worth knowing. Deletion is real: a removed record leaves the current file, and what you get back is whatever git history holds — so your backups are your git history, and keeping it is your responsibility. And an AI with a memory string can write and delete like you can; a memory string is a key, so treat it like one.

This is the plain-language version of the warranty and liability sections of the Apache License 2.0, under which Nodus is released. Where they differ, the licence text governs.

Who provides this

Nodus is made by BeyondEnd (Erhan Olgunelma). The hosted engine runs at nodus.beyondend.dev. The source is public at github.com/olerhan/nodus, so everything described here can be checked rather than taken on trust.

What the engine keeps

Not your notebook. Your records live in the git repository you point Nodus at. On each request the engine fetches the notebook, builds an index in memory, does the work, writes back if anything changed, and forgets it. No volume is attached to the container, and there is no database of user data.

WhatWhere it livesHow long
Your records Your git repository Until you delete them; git history keeps prior versions
Files you attach Your own device, under the node's id Until you delete them
Notebook contents on the engine Process memory only The length of one request
Git installation tokens A small bounded in-process cache Up to one hour, and lost on restart
Rate-limit counters Process memory only The length of the rate window

The engine keeps no account, no user list and no profile, because it has no notion of a user: whoever holds a memory string opens that memory.

What we do not do

We do not sell your data, we do not share it with third parties for advertising, and we do not train models on it. We do not read your notebook: it is not stored anywhere we could read it from.

Your notebook does travel to the parties you choose, and it is worth being clear about who they are: your git host (GitHub, Cursor Origin, or wherever your repository lives) and the AI vendor you connect, whose own policies apply to whatever passes through their model. Nodus adds nobody to that list.

Logs

The hosting platform records ordinary operational information for the service to run — request times, status codes, IP addresses at the network layer. Memory strings and notebook contents are never written to a log.

Authorising a connector

When you connect an AI, the engine uses OAuth 2.1 with PKCE. The access token it issues is your memory string, so the connector holds the address and key to your repository and nothing else is stored on our side. Revoking access means rotating the key at your git host or reconnecting with a new string.

Where a GitHub App installation is involved, the app's private key is sealed with an encryption key held only by the engine process. It is used to mint short-lived installation tokens and is never returned to a client.

Children

Nodus is not directed at children and is not intended for use by anyone under 13.

Changes and contact

If this policy changes, the updated version appears on this page. For questions about privacy, security or anything on this page, write to team@beyondend.dev. Security reports are welcome and answered.

Last updated: 23 August 2026.