Your company inbox, in the same place as everything it refers to — so a thread is never missing the file, the task or the call it belongs to.
AI Business OS
AI is your most important agent now.
Your responsibility is to give it the right context for the right task — so reorganise the business around it, starting now. Majig unifies projects, tasks, email, files, notes, chat and meetings around your company's vision and targets, with agents in every room and the latest AI models always at your back.
Start now — early access, onboarded by hand, a few businesses at a time.
Your business already produces everything an AI would need. It just throws it away.
The decision was made on a call nobody recorded. The reason sits in an email thread one person can find. The process lives in a chat message that scrolled past in March. Every day your company generates exactly the context that would make AI useful to it — and every day that context evaporates into five disconnected tools.
Majig keeps it. Not by asking anyone to write things down, but by being the place the work already happens — and having agents in the room while it does.
One platform
The places work happens, under one roof.
Not another tool beside the others — the replacement for the gaps between them. Everything in one system means everything shares one memory.
Chat
Direct messages, team channels and per-task threads. The conversation sits next to the work it is about, not in a separate app.
Meetings
Internal calls and external guests, hosted from the platform. Which matters because a meeting held here is a meeting the business remembers.
Files
Documents attached to the entity they describe — a contract to a supplier, a spec to a project — not buried in a folder tree nobody maintains.
Notes
The half-formed things: a price someone quoted, a reason you chose a supplier. Written once, found forever, linked to what they are about.
Calendar & tasks
What is due and who is doing it, sitting in the same memory as the conversation that created it.
Always in the room
Agents that attend everything.
Not a chatbot you remember to open. Colleagues who are present in every channel, continuously, and who are still paying attention on the Friday afternoon when nobody else is.
Every email
Who promised what, to whom, and by when
Every chat message
The decision that happened in passing, captured
Every meeting
Internal and with guests — what was agreed, not just that it occurred
Every file and note
Read, understood, and filed against what it concerns
Every task and date
Tied back to the conversation that caused it
A person who joined your company today would need a year to learn what your business knows. An agent that has attended every conversation since you started already knows it — and unlike the person, it never leaves, never forgets, and is in all the rooms at once.
That is the whole bet. Attendance is the moat. Models will keep getting better and cheaper; what they will not have is the context only your company can produce.
The business brain
Context in. Structure out.
Everything the agents hear is distilled into an organised record of your business — entities, processes, people, decisions — embedded and indexed as it is written.
Monday, 14:06 — a call nobody would have written up
Agent attending
In the meeting
"We're switching packaging to Metin in Antalya — 30-day terms. First order Monday. Keep the old supplier as backup until we've seen two clean deliveries."
Filed as a new supplier, a payment term, a task and a decision with its reason. The Antalya address already existed and was referenced, not copied.
entities/suppliers/metin-packaging.mdNew
entities/suppliers/↳ terms: net-30New
decisions/packaging-switch.mdReasoned
processes/procurement.mdUpdated
tasks/first-order-monday.mdNew
entities/places/antalya.mdReferenced
Every change is an event you can read, undo, or ask about later.
Illustrative
- Embedded on write. Every record is turned into vectors as it is filed — ask "who did we use for packaging in Turkey" and it finds the answer even though you never wrote those words.
- Chunked so meaning survives. Long documents are split at sensible boundaries, so a match points at the exact passage rather than a whole file.
- Multilingual by design. Ask in one language about something said in another. The embedding space does not care which.
- It walks, it does not load. A small index is always in context; everything else is pulled only when the question needs it — so ten thousand records answer as fast as ten.
Always loaded
Index
A small, cached map of what exists and how to reach it.
~1–2kOn demand
Area documents
Procurement, customers, finance — opened only when relevant.
pulledOn demand
Knowledge & records
The exact file the question needs, found by vector search.
matched
AI switchboard
Connected to every AI — and honest about which one to use.
Majig is not one model in a box. It is the place your business connects to all of them, with each job sent to whatever currently does it best or cheapest.
Reasoning
Hard analysis, planning, judgement
→ frontier model
Volume
Triage, tagging, summarising at scale
→ small fast model
Speech
Transcription and voice
→ specialist service
Vision & documents
Scans, invoices, photos
→ purpose-built model
The market moves every month. A task that needed the most expensive model in spring runs at a fraction of the cost by autumn. We follow that market so you do not have to, and move your work onto whatever is now the efficient choice.
And it extends. New model, new provider, your own internal service — point Majig at it and the business brain behind it comes along unchanged. You are never buying into one vendor's idea of the future.
Your infrastructure
Your server. Your accounts. Your keys.
Majig is not a pool you rent a corner of. Each business runs on its own machine, under its own cloud account, with its own credentials — set up for you, owned by you.
A server of your own
Your business brain lives on your instance, not in a shared tenant. We provision and run it; the data never mingles with anyone else's.
Your cloud, your subscriptions
Services are bought in your name, on your billing. If you ever part ways with us, the accounts and everything in them stay yours.
Credentials in your own vault
Every key and password sits in your 1Password vaults — referenced by the system, never absorbed into it. You can revoke any of it, any time, without asking us.
We bring your past in
Your existing mail and conversations are migrated in during onboarding, so the brain starts with your history rather than an empty page.
How we charge
One flat fee. And AI below list price.
Most AI products resell tokens at a markup above list — which quietly rewards them for making the system chattier. We run it the other way round: our job is to make your AI cheaper than you could buy it yourself.
$190/ month
or $1,900 / year — two months free
A flat fee for the platform
Runs on your own server, on your own account's tokens. One predictable charge for the software, the setup and keeping it running — it does not move because an agent had a busy week.
- The unified workspace — mail, chat, meetings, files, notes
- Provisioning and running your instance
- Migrating your existing history in
- Keeping you on the most efficient models as the market shifts
Your accounts — or our rates. Your choice.
Your own provider subscriptions and API accounts work out of the box; nothing is locked to us. Or opt in to the Majig channel: usage pooled across all our businesses, with rates negotiated below public list prices at every major AI service.
- Opt-in, never forced — switch between your accounts and ours at any time
- Below list when you use our rates — pooled volume is a discount you cannot get alone
- Our commission lives inside the negotiated spread — never on top of your bill
- Routing each job to the most efficient capable model cuts your bill again
The alignment, spelled out: we only earn when you are paying less than you would without us. And the more businesses run on Majig, the better the rates we can negotiate for everyone already here — loyalty literally compounds into discount.
Why start now
The AI you will run in three years needs data you have to start keeping today.
Every company will eventually want AI deep inside its operations. The ones that get there will not be the ones that bought the cleverest model — they will be the ones whose last few years of work are sitting in a structure a machine can actually read.
That archive cannot be bought later. You cannot retroactively attend the meetings you already held or recover the reasoning behind decisions nobody wrote down. The only way to have it is to have been keeping it. Majig starts keeping it from the day you turn it on — AI-native, machine-learning-ready, and yours in plain text if you ever walk away.
And it has a destination. Once the brain holds enough of your business, the next step stops being a better prompt and becomes a model shaped by your own data — trained on how your company actually works. That is only possible for a business that started keeping it.
Team & trust
Agents hear everything. That only works if the rules do too.
A system this close to your business has to be trustworthy by construction, not by promise.
Everyone works on the same record
When one person tells Majig something, the business knows it — not just that person's chat window.
- People and agents, one permission model. Every agent has a file, a registry entry and its own permissions, exactly like a colleague.
- Shared projects and threads. Work has a home that outlives the conversation.
- Every change has an author. Who changed what, when and why — for people and agents alike.
Private by default, shared on purpose
Attendance is only acceptable if access is explicit and nothing acts irreversibly on its own.
- Explicit visibility. Private or internal by default; anything wider is a deliberate grant, never an accident of search.
- Secrets stay in a vault. Keys and passwords are referenced, never stored in the record.
- Approval gates that cannot be earned away. Money, credentials and anything leaving the building ask first, every time.
- Plain text, portable. Readable by you and by git. You can always take it and leave.
Early access
Start keeping what your business already knows.
We onboard a small number of businesses at a time, by hand — because the first weeks are spent teaching the system the shape of yours.