Describe the work your business needs done. Ultron OS builds the team that plans and does it, and you decide which actions wait for your approval and what each agent may spend.
You describe the work. The agents plan it and do it, and stop where you asked them to.
Describe what a department should achieve and press Build department. Ultron OS designs the roles and hires the whole team in one step.
The agents work through tasks. On the ones that matter, a second agent reviews the work before it moves on.
Human-in-the-loop approvals: pick which kinds of action need your OK. They wait in your Inbox until someone on your team decides.
Give the agents the apps your business already uses. Find one by name or web address, sign in and choose which agents may use it.
Give each agent a monthly spending limit. When it reaches the limit, that agent pauses until you resume it.
Bring your own key from the AI provider you already use. The provider bills you directly for what the agents use.
Your agents and their work live in a workspace of your own, with its own database, apart from other customers.
Home shows each department and its agents, and tells you how many tasks are waiting for your decision.
Every department, its agents and what they are doing now, with all tasks in one list.
Each open task drawn next to the agent doing it. Pick an agent to give it something new.
The built-in assistant. Ask, instruct or jump to a page. It shows which actions happen at once and which need approval.
Speak to Ultron once you connect a voice provider. Live conversation needs a provider that supports it.
Your company knowledge base, drawn as notes and their links. Every agent can search it and file notes into it.
Everything your agents have produced, across every department, in one place.
A department's plans, tasks, team, automations, output and what waits for you.
We set up your workspace, you connect your AI provider and describe the work. The team is built from that.
Describe the work, and Ultron OS hires a team of AI agents to do it. You set what they may do alone and spend.
Give it a mission, a goal and a size, and the whole team is hired at once.
Each agent gets a role and its own instructions. A lead plans and hands out the work.
A monthly team budget is split into a limit per agent.
Hire more later
Add ready-made roles as you grow. New hires start paused.
Each task has a brief, a priority and an agent who owns it.
Name a second agent to review or approve important tasks. The task waits for its decision.
From a team, an agent, the team map, or Ultron.
A log per task
See its status, who worked on it and what they did.
Choose which actions wait for your OK, and they queue in the Inbox until someone decides.
One limit: messages your agents send through a connected app don't wait for approval. Responsible AI lists the others.
Upload your documents once, and every agent can search them.
Upload files, save a chat attachment, or import a whole folder.
Every agent can search all of it, so keep private files out. Found text reaches agents as reference material, a first defence against hidden instructions.
Brain
Your notes and their links. Agents add their own.
Type an app's name or address, and Ultron OS works out how to connect it.
Your AI provider bills you for what agents use, and we charge the plan price.
This month, the projected total, and cost by team, agent, job and model.
At its monthly limit, an agent pauses until you resume it. Changes take a few minutes.
Each agent has its own model, the strongest going to the lead.
Run a team on a schedule, or when something happens in another app.
You approve it once, when you turn it on. Then it runs without asking again.
Plans
Basic and up. On Free, scheduled runs are cancelled.
No. We host it and set up your account for you.
Yes, and your provider bills you. Anthropic, OpenAI, Google, xAI, OpenRouter or your own Anthropic-compatible endpoint all work.
The request stays in the Inbox. It never expires.
We run everything for you, and your plan decides whether you share a server.
Each workspace runs in its own environment, separated from every other customer at the process, database, user and network level.
Each workspace runs its own engine process under a dedicated operating-system user, with its own PostgreSQL instance (separate data directory and port) and its own firewall rule. The engine's data for your agents lives in that dedicated database.
Sign-in, sessions and the credential vault sit in one shared account database. Shared operational tables carry a workspace identifier that scopes each row.
AI provider keys are encrypted with AES-256-GCM under a master key held per environment. App credentials are attached server-side just before each call and never enter an agent's context.
AES-256-GCM
Encryption for stored AI provider keys.
Staff access
Only by an explicit, revocable per-person grant.
Leaving
Removing a workspace removes its data store.
Your plan decides whether other customers share your server.
A shared server. Each workspace keeps its own process, database and user, capped at 2 GB of memory.
A dedicated virtual server that runs your workspace and no other customer's.
A larger dedicated virtual server, for up to 12 agents at once.
Connector calls leave through a single guarded client, and MCP calls go only to the app's own host.
In the European Union, on servers we run. You cannot pick the cloud or region, or install it yourself.
Only a staff member holding an explicit, per-person grant, which can be revoked. Anything unrecognised grants nothing.
No. Support times are set per plan.
Approvals and spending limits keep you in charge of your agents. Here is where each one stops.
Actions you set to wait go to the Inbox, but three kinds never do.
Messages your agents send through a connected app don't wait for approval. Give each app only to the agents that need it.
You approve an automation when you turn it on. After that it runs by itself.
The Inbox keeps recent decisions and their outcome. It does not record who made them or why.
Three more safeguards, each with one caveat.
An agent pauses at its monthly limit. Team and account budgets are guides only, and a changed limit takes minutes to apply.
Agents reach apps over secure, checked addresses. An app you add by hand can reach your own network.
What agents read in your knowledge base reaches them as reference material, a first defence against hidden instructions. Every agent can search all of it.
Straight to the AI provider you connected, on your own account. The provider bills you for it.
A task's reviewer and approver are agents. Only the Inbox waits for a person.
Yes. Resume it, or raise its limit.
Hand a team of AI agents the work that eats your week. Sales and Marketing come ready; build others from a goal.
A ready-made Sales team researches each new lead and drafts a reply.
Example mission
Look up each new lead and draft our first reply.
A ready-made Marketing team drafts on a schedule you set.
Example mission
Two posts a week from our product notes, in our house style.
Build a support team that watches your helpdesk and drafts replies.
Example mission
Draft answers to simple tickets from our help pages.
Build an HR team that answers from your handbook and onboards new starters.
Example mission
Answer handbook questions and onboard each new starter.
Build a team that reads each bid against your criteria.
Example mission
List what each submission leaves out of our published criteria.
Build a team that checks each contract against your positions.
Example mission
Flag each clause that differs from our standard terms.
Agencies, sales-led firms, logistics teams and professional services, with one person running operations.
You can name the work to hand over and what needs your OK.
Consumers, enterprise buying cycles longer than six months, and regulated decisions made by an agent.
Set once for every team; Responsible AI shows where each stops.
The actions you picked wait for your OK.
Every team searches the same documents.
Each agent pauses at its monthly limit.
Pay for how many agents work at once. Prices include VAT, and runs are unlimited.
€0/ month
Shared server
€49/ month
Shared server
Everything in Free, plus
€129/ month
Private server
Everything in Basic, plus
€179/ month
Private server
Everything in Pro, plus
All four include unlimited runs, a private workspace and a spending limit per agent.
| Free | Basic | Pro | Ultra | |
|---|---|---|---|---|
| Price | €0 | €49 a month | €129 a month | €179 a month |
| Agents at once | 1 | 2 | 6 | 12 |
| Departments | 1 | 2 | 6 | Unlimited |
| Server | Shared | Shared | Private | Private |
| Automations | No | Yes | Yes | Yes |
| Support | Community | Email, two business days | Priority, same day | Named contact |
Your AI provider bills you for what agents use, and we bill the plan.
Yes. Free costs €0, with one agent working at a time on a shared server.
How many agents can work at one moment. On Basic, two work while others queue.
The agent pauses until you resume it or raise its limit.
Tell us the work that stalls first and the plan you have in mind.
We agree a plan with you, then open your account from your email and a name for it. You sign in with a one-time password and choose your own. Then you connect your AI account and build your first team from a goal. The setup guide walks through each step.
Your answers pick the plan and shape your first team.
One repetitive job: how it runs today, who touches it and where it gets stuck.
What should wait for your OK, which apps each agent may use, and what each may spend in a month.
How many agents need to work at once, and whether a shared server is fine for you.
None of these is required, but each one saves a step.
What Ultron OS does, how it differs from a chatbot, and the words you’ll see on screen.
What it does, what you control, and what you bring along.
It hires a team of AI agents from a goal you set, and tracks their work as tasks you can open and follow.
What waits for your OK, which apps each agent uses, and what each agent may spend a month.
An AI account, one job to hand over, and someone to answer approvals. Setup covers the details.
A chatbot waits for your next message, while a team works through tasks and brings you what needs a decision.
| A chatbot | A team in Ultron OS | |
|---|---|---|
| Starts when | You write | A task, a schedule, or something happening in another app |
| Unit of work | A reply | A task with an owner and a status |
| Who does the work | You, with the reply in hand | An agent in a role, with its own spending limit |
| What waits for you | Your next message | Approvals and questions in your Inbox |
Three steps from your first email to a working team.
We open your account, and you build the team.
From your email and a name, on the server your plan includes.
Use the one-time password we send, then connect your AI account.
Give it a name, mission, goal and size, then press Build department.
The whole team arrives at once, with instructions for each role.
Decide these before the first task starts.
After launch, these are just settings.
'%20stop-opacity%3D'.22'%2F%3E%20%3Cstop%20offset%3D'.36'%20stop-color%3D'rgb(63%2C191%2C166)'%20stop-opacity%3D'.09'%2F%3E%20%3Cstop%20offset%3D'.70'%20stop-color%3D'rgb(23%2C74%2C66)'%20stop-opacity%3D'.035'%2F%3E%20%3Cstop%20offset%3D'1'%20stop-color%3D'rgb(15%2C17%2C19)'%20stop-opacity%3D'0'%2F%3E%20%3C%2FradialGradient%3E%20%3CradialGradient%20id%3D'core'%20cx%3D'50%25'%20cy%3D'50%25'%20r%3D'50%25'%3E%20%3Cstop%20offset%3D'.52'%20stop-color%3D'rgb(10%2C12%2C14)'%20stop-opacity%3D'.88'%2F%3E%20%3Cstop%20offset%3D'1'%20stop-color%3D'rgb(10%2C12%2C14)'%20stop-opacity%3D'0'%2F%3E%20%3C%2FradialGradient%3E%20%3C%2Fdefs%3E%20%3Crect%20width%3D'800'%20height%3D'800'%20fill%3D'url(%2523w)'%2F%3E%20%3Ccircle%20cx%3D'400'%20cy%3D'400'%20r%3D'236'%20fill%3D'url(%2523core)'%2F%3E%20%3Cg%20fill%3D'none'%20stroke%3D'rgb(125%2C134%2C131)'%3E%20%3Ccircle%20cx%3D'400'%20cy%3D'400'%20r%3D'168'%20stroke-opacity%3D'.30'%2F%3E%20%3Ccircle%20cx%3D'400'%20cy%3D'400'%20r%3D'232'%20stroke-opacity%3D'.18'%2F%3E%20%3Ccircle%20cx%3D'400'%20cy%3D'400'%20r%3D'300'%20stroke-opacity%3D'.10'%2F%3E%20%3C%2Fg%3E%20%3Cg%20fill%3D'none'%20stroke%3D'rgb(87%2C217%2C192)'%20stroke-opacity%3D'.55'%20stroke-width%3D'2'%20stroke-linecap%3D'round'%3E%20%3Cpath%20d%3D'M400%20168%20A232%20232%200%200%201%20564%20236'%2F%3E%20%3Cpath%20d%3D'M400%20632%20A232%20232%200%200%201%20236%20564'%2F%3E%20%3C%2Fg%3E%20%3C%2Fsvg%3E)