The agent that sees, hears and acts with you in real time.
Saelyx sees your screen, sees you through the camera and hears you—all three at once—then reasons over everything it takes in and acts in about a second. An agent that works beside you, live.
All three inputs, at once.
Other agents do one thing at a time. Saelyx sees your screen, sees you through the camera and hears you simultaneously, and reasons over all three together—in real time, with ~1 s latency.
Screen
Reads every pixel while you work.
Camera
Sees you, and whatever you hold up to it.
Voice
Hears you and answers out loud.
One agent · live
Three input streams · one line of reasoning · an answer in ~1 s on a real-time engine
The more it knows you, the better it works with you.
Memory is the moat.
Saelyx remembers names, preferences and context, and searches inside your own memories. Next time you don't start from scratch: it already knows who you are and what matters to you.
What I remember
It sees what you show it.
Share your screen or point the camera at something. It recognizes objects, faces and scenes, and answers out loud. The conversation is genuinely multimodal: sight, voice and screen at the same time.
Turns you can see.
You always know whose turn it is. The orb cools down when it thinks and warms up when it speaks. No talking over you, no awkward silences.
- Listening — bright, receptive
- Thinking — dimmed, no ripples
- Speaking — warm, radiating
On every device you use.
Available on Mac today, and you can already talk to it by voice in the browser. iPhone, iPad and the full web version arrive in Q4 2026, Android in 2027. Start on one device and carry on in another without losing the thread—or the memory.
Everything Saelyx does.
One agent, many hands. This is what it can do with you, live, on in-house technology.
Sees your screen, sees you through the camera and hears you—at the same time.
All 3 inputs at once
Screen, camera and microphone at the same time. One line of reasoning over everything that happens, in real time.
Screen vision
Sees every pixel while you work and understands what's in front of you.
Live camera
Show it something and it understands right away, as if it were holding it.
Real listening, real voice
Natural conversation: 776 ms to the first syllable, with a voice you can interrupt.
Keeps what matters and goes back to it when it's needed.
Persistent memory
Remembers names, context and preferences between sessions, and follows you from one device to the next.
The memory is a plain list of sentences: you can read it, edit it and delete it.
Knowledge base
Upload your documents and it answers citing them, with the exact source.
It answers citing the document and the exact passage, not a blurred version of it.
Face recognition
Builds a searchable visual memory: who is who, and what was going on.
Only if you turn it on, and faces never leave the scope of your account.
Writes, runs and produces for you, inside your own apps.
Silent Meeting Mode
It listens to the meeting and feeds you the exact line in a panel that floats over the call. Clicks pass straight through it.
What you see in the panel is not the transcript: it's what Saelyx suggests you say. The SAY THIS tags mark the exact line; the rest is the data nobody has put on the table yet.
Dictation · Whisper Mode
Types for you in any app, turning your voice into text as you go.
Terminal by voice
Runs shell commands and drives interactive CLI tools on your Mac.
It's your real shell, inside Saelyx: the commands actually run and the output is the real one. You can talk to it instead of typing.
Generates documents
PDF, Word, Excel and PowerPoint dictated by voice, ready to send.
You dictate it and the file comes out finished, ready to send.
12 specialist roles
Engineer, Lawyer, Doctor, Financial analyst, Teacher... switch roles mid-conversation.
You switch roles mid-conversation and it keeps your memory and your context.
On all your devices, and without keeping your data.
Multi-device
Mac today; iPhone and iPad in Q4 2026. The same session and the same memory, in sync.
The same session follows you from the Mac to the iPhone to the watch.
Zero Data Retention
Where it applies, audio and screen are processed on the fly and discarded: they are not stored and are not used to train anything. The exact scope is in the privacy policy.
How it works.
Talk.
Press once and talk the way you normally would. Saelyx hears you in real time and answers out loud, with almost no wait.
Show it.
Share your screen or your camera. It understands documents, faces and scenes, and reasons about what it sees.
It remembers.
What matters stays. It builds a memory of your own—names, context, documents—and goes back to it when it's needed.
Watch it work.
You talk to it and it answers out loud, while it's happening. You don't wait until the meeting is over.
Recorded in the app, no cuts, no editing. The conversation is an example.
Why we built it this way.
We think an AI agent should feel like someone who knows you, not like a form that talks. That's why Saelyx sees, hears and remembers at the same time. That's why it knows when to act and when to keep quiet.
We built it because we were tired of waiting until the end of the meeting for the machine to understand what had already happened. Saelyx lives in real time: with you, while it happens, not afterwards.
In-house technology, and Zero Data Retention where it applies: there your audio and your screen are processed on the fly and discarded, never stored and never used to train anything. The exact scope, with no small print, is in the privacy policy. Built to last, designed to feel close.
— Adianny & Karina · A Coruña · 2026
Meet Saelyx.
Ask for access. I review every request by hand and write back as soon as there's room for you.