The second brain that is everywhere you work
Everything you read, hear and save — compiled into one connected memory that remembers for you, and is there wherever you work.
Your knowledge is everywhere. Your memory isn't.
You already have good places for what you know — a notes app, a drive full of documents, bookmarks, a wiki, years of mail. They are excellent at keeping things. What almost none of them do is work on what they keep: every file stays a file, and every answer is still put together by you, one search at a time.
Trail adds the step those tools leave out. It compiles what you give it: each source is read once and turned into connected, citable knowledge, linked to what you already know and flagged where it agrees or disagrees. So when you ask, the answer is already assembled — with the sources it came from.
That is what Trail is. One brain for everything you know — and it is there wherever you work.
One brain, many doors
A second brain is only useful if feeding it costs nothing. So Trail does not ask you to move into a new app. It meets you where the knowledge already is.
Drop in a file
PDFs, Word documents, PowerPoint decks, Excel sheets, images, audio recordings and markdown all go in the same way. Trail reads them and keeps the original as the source.
Clip from the browser
The Trail Web Clipper saves the page you are reading straight into your brain with one click. It works in Chrome, Edge, Brave and Arc, and stays idle until you click it — it never follows your browsing. (Available as a download today; the Chrome Web Store listing is in review.)
Put it on your website
<trail-chat> is a ready-made chat window for any web page: one script tag and one HTML element. Your visitors ask in their own words and get answers from your brain, with citations to the sources. It reaches Trail through a small relay on your own server, so your API key never ends up in a visitor's browser. How to set it up.
Wire it into your tools
A REST API lets your own apps and AI agents search your brain, chat with it, and send new knowledge in.
Talk to it
Chat with your brain in the Trail app and every answer comes back with citations to the sources it was built from.
Many doors, one brain. Whatever comes in through any of them ends up in the same place, connected to everything else.
It doesn't just store. It compiles.
Most "chat with your documents" tools keep your files as a heap of text fragments and search through them every time you ask a question. Ask the same thing twice and they do all the work twice. Nothing is ever learned.
Trail works the other way round. When a source comes in, Trail compiles it — once — into Neurons: short, citable notes that each hold one piece of knowledge. Neurons link to each other, so a new document does not just sit next to the old ones. It joins them, and you can see what it connects to, what it builds on and what it contradicts.
Every Neuron remembers where it came from. Click through and you are back at the original PDF, page or recording.
The result is a brain that gets more useful the more you put in, instead of more cluttered. We wrote the full argument in Compile-Time Knowledge: A Technical Argument Against RAG.
It gets better, not just bigger
A brain that believes everything it is told is not much of a brain. Trail has a gate.
New knowledge lands in a curation queue before it becomes part of your brain. Clear-cut additions can be approved automatically; anything that needs a human eye waits for you. Trail also keeps checking what is already there — flagging Neurons that contradict each other, notes that have gone stale, and orphans that nothing links to — so the brain stays something you can trust.
How that gate works, and why it is modelled on how human memory decides what to keep, is in Three Filters on the Gate.
Yours, and kept in Europe
A second brain holds the things you actually know, so where it lives matters.
Stored in Stockholm
The Trail app, the engine and the databases all run in Stockholm, Sweden. Each customer's brain has its own database — it is never mixed with anyone else's.
EU models in the cloud
When Trail compiles your sources and answers your questions in the cloud, it uses Mistral, a French AI provider, in the EU. If that provider is down, Trail fails visibly rather than quietly sending your data to a model outside the EU.
Every source kept
Your originals stay attached to the knowledge built from them, so you can always check what Trail has learned against what you gave it.
Start your second brain
Trail is the memory machine Vannevar Bush imagined in 1945 — we wrote about that, too — built for the way you work now: everywhere, all at once.