Use cases · live demos
Ten working demos — open one, turn on your camera, and watch an agent perceive the real world. Each has an Afferens toggle: raw guess vs verified read. Runs in your browser, on phone too.
Vision is live today. Spatial, acoustic and more are rolling out.
Confirm the target before you move.
Point at a cup. The agent locks the target — but only acts when the read is verified.
Point a camera, get the shelf as data.
Pan across items — the agent builds a verified, structured count of what it sees.
Hold until it reads clear.
Detects a person in the zone and refuses to green-light until the read is verified.
Describe the scene out loud.
The agent perceives the room and speaks what it sees — accessibility, hands-free.
Flag and log with evidence.
Watches for a target item on the line and logs a verified pass/fail.
Identify the device, guide the fix.
Recognises the device in front of you and overlays the next step.
Reacts to who it sees.
Sees a person approach and greets them. (Voice layer is roadmap; vision is live.)
Snap an item, it’s logged.
Point at objects — verified ones auto-log to stock.
Read presence and intent.
Real-time hand tracking — reads the gesture, not just the pixels.
If the agent has to actually see — that’s the bar.
Anything where grounding an agent in the physical world is the hard part.
One API call. Vision live now. 10,000 free tokens, no card.