A home assistive
Robot
you can teach.

Teach AhaRo skills and routines in your own home or facility by talking, showing, or pointing. When it is unsure, it asks before acting, and it remembers what you teach it.

AhaRo: a slim wheeled base, a tall column carrying two small arms with grippers, and a camera head on top. One arm reaches forward.
You might ask
  • Bring me my glasses.
  • I spilled my tea. Can you help?
  • Where did I put the remote?
  • Did I take my pills today?
  • Did anyone leave a light on or a window open?
  • Can you water the plant in office 3.30?
  • Can you turn off the light in the office?
  • A little bigger than a robot vacuum cleaner
  • Weighs less than a toddler
  • Developed in Europe by Intellectus Agens

Your robot does not need to work everywhere.
It needs to work at your place.

Instead of learning a general solution, we provide a general learning method for user-specific solutions. You teach AhaRo your home or facility and the tasks that matter to you.

Generalist robot approach

  • designed for broad environments
  • large-scale pretraining
  • keeps going when things change
  • predefined task sequences
  • generic task execution

AhaRo approach

  • adapts perfectly to your specific home or facility
  • requires only a few demonstrations from you
  • asks when confidence is low
  • composes learned skills for novel requests
  • your preferences, remembered

Show it once.

All in plain language: no app, nothing to program. You talk to AhaRo as you would to a helper, and it talks back. Teaching is not setup. It is everyday use, and it gets shorter each time.

firm pull, then it swings a bit higher this one? fridge: firm pull cups: on the shelf cloth: third door 1bottle 2fridge 3door
YouThis is my home. Here is the fridge, and this is how you open it.

Skills it has learned

  • cups, bottles, cloths, whatever ends up on the floor
  • stable placement on a shelf, a table or in the sink, without tipping
  • the cloth, the fruit bowl, your glasses, wherever they were last
  • the fridge needs a firm pull, the bedroom door swings, the balcony door slides
  • on its own, when the battery runs low

New requests, no new teaching

  • find the cloth, bring it, take the cup to the sink
  • open the fridge, pick up the juice, open the balcony door, bring it out
  • pick up the cups, place them by the sink, dock
  • find it where it saw it last, pick it up, bring it
  • checks what it saw this morning and answers. And at 13:00, unasked: opens the drawer, picks up the pills, brings them to you
  • room by room: looks at every window and light, switches off what it can reach, closes the doors, and sends a note of what it found

Built to be next to humans.

Light enough to lift into a car, slim enough for any doorway. Sized for a flat or an office.

Front view of AhaRo with its camera head, two arms mounted on a carriage on the central column, and the wheeled base
  • A depth camera sees the room, the objects in it, and your gestures. When you point, it works out what you mean.

  • Lightweight arms with two-finger grippers, sized for cups, bottles, handles and switches. They give way if they meet something unexpected.

  • Both arms ride the column, so AhaRo reaches a remote on the floor and a cup on the counter.

  • A little bigger than a robot vacuum. Carries the computer and battery on two drive wheels and casters, low to the ground.

What it does,
at home or in a facility.

Everyday abilities to start with, more as you teach it. You ask in plain language; it answers the same way.

  • Brings what you name or point at. Bring me my glasses.

  • Everyday speech. Asks back when unclear, tells you what it is about to do.

  • Learns your layout and moves between rooms on its own.

  • Picks up, places, tidies, puts away.

  • It goes and looks, then tells you. Where did I leave the keys?

  • Medication, meals, appointments. Brings what the reminder is about.

  • Swinging or sliding, once it knows which is which.

  • Which pot, how much, how often, once you have shown it.

  • Fills the bowl on schedule, throws the ball when you are out.

  • Ask from anywhere, it goes to look. Is the stove off? Or on a schedule: doors, rooms and equipment, with a note of what changed.

  • Docks on its own when the battery runs low.

  • Someone you choose sees what was done, adjusts the schedule, and steps in from their phone when asked.

Who it is for.

  • An extra pair of hands for reaching, carrying and remembering, without giving up your own place. In the first pilots, some households will try AhaRo remotely and a few will host it at home, with our engineers alongside. Leave your email

  • There between your visits, and tells you what happened. Tell us about your relative's home and routine, and who should be kept in the loop. Leave your email

  • Bowl filled, ball thrown, plants watered, on schedule, and the home looked after while you are away. A good first case for a remote trial. Leave your email

  • Taught once per resident's room and routine; takes the small errands off staff rounds. We are looking for a few providers in Vienna to run AhaRo in a resident's room or on a care round, with our team on site. Leave your email

  • Taught once per building and round: doors, rooms and equipment checked on a schedule, small errands on the way, and a note of what changed. We are looking for a few sites in Vienna to run AhaRo on a round, with our team on site. Leave your email

From passive to active intellect.

AhaRo knows what it has learned, how confident it is, and when it needs help. Today's robot AI recognises and imitates well, but it does not know when it is unsure and does not plan past the next step. AhaRo adds that proactive part, built on trustworthy AI methods.

For technical readers

Where most robot AI stops

  • it has read a lot
  • it can tell a cup from a bottle
  • it can copy a movement
  • it has a rough idea how things move

Where AhaRo starts

  • it knows when it is unsure
  • it knows when to hand over to you
  • it learns from you as you go
  • it thinks a task through before acting

How it thinks

A hybrid cognitive architecture, neuro-symbolic in research terms: learned skills for the physical world, a step-by-step planner that chains them into longer sequences, and a memory of your home or facility. The memory holds where things are, how they work and how you like them handled.

How it works with humans

An autonomy arbitration layer decides, moment by moment, who holds control: the robot, you, or a trusted person connected remotely. Speech and gestures run both ways. It announces what it is doing, pauses for confirmation below the acceptable threshold, and offers help when it recognises a routine. It learns your preferences from what you say and correct, occasionally asks how satisfied you are, and adjusts.

How it stays safe

Every skill carries a calibrated confidence estimate; below the acceptable threshold AhaRo stops and asks you or a trusted person. Alongside learned skills it keeps symbolic safety rules and a model of acceptable behaviour, preprogrammed and extendable through corrections by a responsible user: rooms it must not enter, hours it must stay put, objects it must never touch. Every plan is checked against them before execution.

How it keeps your privacy

Locally hosted AI: the models run on the robot, and what it learns about your home or facility stays there. Only selected, anonymised data leaves the building, with your consent; sharing with a relative, carer or colleague is limited to the people you choose. The information flow is transparent: you can see what it has learned, remembers and shares.

The research behind AhaRo

  • Ajanović, Prakash, de Souza Rosa, Kober. IEEE Robotics & Automation Magazine, vol. 33, no. 1, pp. 130–141, 2026.Teaching long tasks step by step, the way AhaRo learns a routine.Read the paper
  • Luijkx, Ajanović, Ferranti, Kober. Transactions on Machine Learning Research, 2025.Knowing when to ask, and asking only then.Read the paper
  • Azirar, Ajanović, Blum. IEEE Robotics and Automation Letters, 2026.Planning long tasks from what the camera sees, without hand-written models.Read the paper
  • Luijkx, Ma, Ajanović, Kober. IEEE International Conference on Robotics and Automation (ICRA), 2026.Using language models to guide a robot toward sensible next steps.Read the paper
  • Ma, Luijkx, Ajanović, Kober. IEEE International Conference on Robotics and Automation (ICRA), 2025.Learning new manipulation tasks faster, from less data.Read the paper
  • Luijkx, Ajanović, Ferranti, Kober. International Conference on Information, Communication and Automation Technologies (ICAT), 2025. First presented at the NeurIPS 2022 Robot Learning Workshop.Spotting when a fetch command is ambiguous, and asking for a demonstration only then.Read the paper

AhaRo is developed by Intellectus Agens in Vienna,
on more than five years of robot learning research at TU Delft and RWTH Aachen.
Key parts of the system are the subject of patent applications.

Pilot with us.

AhaRo is not for sale yet. Intellectus Agens will shape it with pioneer customers in Vienna, around real homes, facilities and routines. Status: working prototype, first pilots agreed.

We write only when there is something real to show, or when you can help.