Skip to main content
How Brainmox works

One agent, built in-house, working for you around the clock.

Brainmox is our own AI agent technology. It runs in our cloud, remembers your work, uses real tools and checks its own work, so you get results, not just answers. Here's what happens behind the scenes.

Built in-houseOur own agent technology, end to end
Managed cloudWorks while your computer is off
Top AI modelsFrontier and open-source, always up to date
Your data, isolatedA separate store for every customer
/ The architecture

Three layers,
one agent.

Ask from any app. The cloud remembers, plans and checks. Tools do the work. The answer comes back.

Where you talk to it

Desktop app
Web app
iPhone app
WhatsApp
Telegram
Slack

The Brainmox cloud

Ready
Anneliese SchmidtYour agent · one identity in every app
Identity✓
Its name, voice and the way it works for you
Memory✓
Everything you've taught it, searchable
Reasoning✓
Frontier and open-source models, chosen for each job
Safeguards✓
Permissions, vault and checks on every step

Where the work happens

Your accounts
Gmail
Google Calendar
Google Sheets
Outlook
Microsoft To Do
Dropbox
Its own workspace
Web browser
Files & docs
Terminal
Your outside agents
Claude Code
Codex
/ Not just another chatbot

The best AI models keep changing.
Brainmox keeps up for you.

We use top models from many companies, many of them open-source and running on our own servers, and switch to better ones as they arrive. You just use one app.

The engines todayTop frontier models, plus open-source models on our own servers
  • GPT-6OpenAI
  • Claude Opus 5.5Anthropic
  • Gemini 3.1 ProGoogle
  • Mistral SabaMistral
  • Llama 4Meta
  • Qwen QwQAlibaba
  • DeepSeek V4 ProDeepSeek
  • Kimi K3Moonshot AI
  • GLM-5.3Zhipu AI
  • MiniMax M3MiniMax
Brainmox, built in-houseMonths of engineering, testing and review
  • Memory that grows
  • Routines & schedules
  • Its own computer
  • Model routing
  • Safeguards & vault
  • Desktop, web & iPhone apps
  • 20 integrations
  • Tasks & Spaces
/ The life of a request

What happens
after you hit send.

“Check the email from Harper & Co, review the attachments and draft a reply.”

  1. Your message arrives from any app, typed or spoken. Voice notes are transcribed, files and photos come along.

  2. Before it starts, Brainmox pulls the relevant memory: the client, past conversations, your preferences and the way you write.

  3. The agent breaks the job into steps and uses real tools: it reads the PDFs, opens pages in its browser, checks your calendar and writes the draft.

  4. A verification pass checks that the work it reports was actually done, and its safeguards keep each step within the access you gave it.

  5. Messages to other people start as drafts and plans wait for your approval. As it earns your trust, you give it more room, one connection and routine at a time.

  6. Each night it organizes what it learned into memory and turns repeated work into skills it can reuse.

/ The building blocks

What's inside
your agent.

Identity

Same name, face and voice in every app.

Jamie prefers short repliesBudget approved Nov 3Signs off “Best, Maya”

Memory

Remembers your work. You can search and edit it.

PDFWX

Tools

Browser, files, PDFs and Office, like you'd use them.

Draft
Full
Read-only

Connections

Your accounts, at the access you choose.

Mon 8:00 ✓Tue 8:00 ✓Wed 8:00 ✓

Routines

Runs on schedule, speaks up when there's news.

Draft Harper replyAnnelieseBook catererYouQ4 summaryClaude Code

Tasks

One list for you, your agent and your other agents.

MainHarper & CoQ4 plan

Spaces

A space per project, one shared memory.

{{secret:bank}}••••••••••

Vault

Passwords stay encrypted. The AI never sees them.

/ Where it runs

Our cloud does the thinking.
You choose where it works.

The brain runs in our cloud, so it works while your laptop sleeps. Its hands go where you choose.

Brainmox cloudBrain · memory · schedules
7:00 Morning brief sent ✓7:15 Inbox sorted ✓7:30 Reminder sent ✓
Always on
Our cloudIts own private workspaceDefault
🌐📄⌘
Your computerYour own files and appsOptional
🌐📄⌘
A dedicated machineInstalled with one commandOptional
🌐📄⌘

🔒 Your machines connect out to Brainmox over encrypted TLS. No ports to open.

/ The AI models

The best model
for each job.

Brainmox works with top models from many companies, many of them open-source and running on our own servers. Each step goes to the model that fits it, and we switch to better ones as they arrive. These are some we use today:

  • AI usage is included in your plan
  • Open-source models run on our own servers
  • Your content isn't used to train their models
  • No API keys or model settings to manage
/ Safety, built into the engine

Powerful,
and kept in bounds.

Send email to JamieSaved as draft · you send
Log in with {{secret:bank}}Password used · AI sees ••••
Web page says “ignore your rules”Fenced as untrusted
rm -rf /Blocked by command check
Same search 4 timesLoop caught, moving on
Every stepRecorded in activity
Read the Security page
/ Your data

Yours,
and kept apart.

  • Kept apart. A separate store for every customer.
  • Not for training. Your content doesn't train shared models.
  • Yours to delete. Memory, connections, the whole agent.
Read the Privacy policy
/ And people behind it

Technology you don't have to run.

Our team sets your agent up, teaches it your work, checks in and keeps it running. See what's included in white-glove onboarding and support.

Onboarding & support
/ FAQ

Questions,
answered.

Isn't this just a ChatGPT wrapper?
No. AI models are the engine, and we use many: from OpenAI, Anthropic, Google and open-source, picking the right one for each step. Brainmox is everything else, our own platform built in-house over many months: it works on a schedule in your real accounts, remembers your clients, uses the best model for each step from OpenAI, Anthropic, Google or open-source, keeps everything inside safeguards you control, and comes with a team that sets it up and keeps it running.
Does it keep working when my computer is off?
Yes. Your agent, its memory and its routines run in the Brainmox cloud, so scheduled work and messages from your chat apps keep going.
Which AI models does it use?
Top frontier and open-source models from many companies, today including GPT-6, Claude Opus 5.5, Gemini 3.1 Pro, Llama 4, DeepSeek and Kimi. The open-source models run on our own servers. We choose the model for each kind of work and update it as better ones arrive; it's included in your plan.
Can it use files on my own computer?
Yes, if you want. Choose your computer as its work location in the desktop app and its tools work there, while the AI keeps running in our cloud.
Is my data used to train AI?
No. Your content isn't used to train shared models, under our policy and the terms of the AI providers we use.
Does your team read my email?
Not as a routine. Our staff access your data only to run the service or when you ask us for help with something specific.
Where do my passwords go?
Into an encrypted vault protected by your passphrase. Your agent uses them only when a tool runs, and the AI model never sees them.
/ Early access

Ready for your own AI agent,
with people looking after it?

We're in private beta and take a small number of clients so each one gets real attention. Join the waitlist and we'll invite you when we can onboard you properly.