NODUS

Your AI deserves a backbone too

Large companies run dozens of departments on a single data backbone: everyone writes to the same records, so there is one version of the truth. Nodus gives you and your AI that same backbone — one memory, two interfaces: you work in a table, it works through tools. And the memory lives in your own repository.

Nodusnod, the knot; us, the mind. Here two minds tie the knot: a person and an AI.

One backbone, two interfaces

How Nodus is put together A person connects from the browser, an AI through its connector; the engine stores nothing and the records live in the user's own git repository. Human interface a table in the browser AI interface connector · nine tools Nodus engine applies the rules · stores nothing reads writes Your git repository notebooks/work/memory.jsonl the single source — records live here
Two faces of the same records: a person looks at a table, an AI at tools. The engine sits in between but holds nothing — shut it down and your records are still in your repository.

Why

The thing you lose most when working with an AI is continuity: every conversation starts from zero, and last week's decision — along with the reason behind it — is gone. Note apps are built for people, databases for developers; in between, there was no place where a person and an AI could write as equals.

Nodus is that place. It isn't a ready-made tracker: you define your own record types — construction progress payments, student attendance, or anything else.

The problem enterprises solved is now yours

What makes big enterprise systems valuable isn't the number of modules: it's that every department sits on the same records. Purchasing and accounting don't write to separate files; they write to the same place, so there is one truth and nobody copies anyone else's spreadsheet. Companies paid millions for that, because the alternative was everyone working from their own copy and nobody trusting anyone.

Today your team has one more member: an AI. And it has exactly that old problem — the chat ends, the memory goes; you keep notes in one place, it keeps summaries in another. Two separate truths.

Nodus builds that backbone for both of you. Same records, two interfaces: you look at a table, the AI at its tools — but you write to one place. “Who said what, and why did we decide this?” finally has a single answer.

The differences are worth stating: setup takes minutes rather than months, no consultants, no licence — and the backbone doesn't sit on a vendor's server, it sits in your own repository.

Three things

The memory is yours

Records live in your own GitHub repository, in a plain-text file. The engine stores nothing; shut it down and your memory is still there.

One model

Job, person, amount, decision, note — all of them are the same kind of “node”. Adding a field means no new table and no new code.

Two interfaces, one truth

You from a table, the AI from nine tools. Same records, at the same time — and every record carries who wrote it.

How it works

  1. Connect GitHub

    Nodus walks you through installing your own GitHub app; you choose which repository it may touch. No password here, no account to create.

  2. Open a notebook

    A notebook is a compartment of your memory: “work”, “personal”, a project… Records in one are invisible in the others, yet all of them live in the same repository.

  3. Give your AI the address

    Add this address to Claude or ChatGPT as a custom connector and approve it once. You never paste a key.

    https://nodus.beyondend.dev/mcp
  4. Start writing

    You write from the table, your AI writes while you talk. A change shows up on an open screen within seconds — and becomes a commit in your repository.

What it is not

Not a cloud Your data never rests with us. The engine attaches to your memory on each request, does its work, and forgets.
No account No sign-up, no password, no session. Your identity stays with GitHub; Nodus only uses the access you granted.
No lock-in The file is in your repository, readable line by line. A tool that has never heard of Nodus can read and write it too.
Not a template You define the record types. The same system looks completely different in two different lines of work.

What is inside the repository

Nothing encrypted or exotic: one record per line, readable text. Git keeps the history — who wrote what, and when.

notebooks/work/memory.jsonl {"id":"k_1","modelId":"tpl_decision","textValue":"Payments go out on Thursday"} {"id":"k_2","modelId":"tpl_note","textValue":"Formwork crew arrives Friday"}