Predictive input
An idea we're exploring to make round-trip feel shorter. Nothing ships yet, and we're not promising when.
✦AI
Built for people. Ready for agents.
Remio already adapts its stream to your network, live — that's engineering, not AI. Its native, low-latency architecture is also a solid foundation for AI agents to use down the road. Nothing about that is shipping yet, and we're not promising a date.
✦Status
Real today. Honest about the rest.
Adaptive streaming quality ships today — plain engineering, not AI. Everything else below is an idea we're exploring, not a feature you can turn on. No dates, no promises.
Ships today · not AI
Adaptive streaming quality
Reads round-trip time in the first second, then picks the bitrate envelope that keeps the picture sharp without spiking lag. LAN or 4G, chosen automatically. The one capability that ships today.
Live bandwidth readH.265 / H.264Host-sideZero config
Read more →
Exploring
Agents
Built for people today. Remio's native, low-latency architecture is a foundation we think AI agents could eventually use to see and act on a real desktop too — nothing ships yet, and we're not promising when.
Native screen captureNative inputEncrypted channel
Read more →
Exploring
Neural upscaling
An idea we're exploring: rebuilding sharper pixels on-device from a lower-bitrate stream. Nothing ships yet, and we're not promising when.
On-deviceClient-side
Read more →
Voice control
An idea on our list: speaking commands to your remote host. Nothing ships yet, and we're not promising when.
Behavioural security
An idea on our list: session anomaly detection that runs on your host. Nothing ships yet, and we're not promising when.
✦Native
Ready for agents, built for people.
Remio's native architecture — direct peer-to-peer transport, hardware-accelerated encode and decode, on-device processing — was built for people first. That same foundation is what makes it a credible platform for AI agents down the road, whenever that day comes.
Direct peer-to-peer transport · hardware-accelerated encode and decode · zero screen data leaves your devices.
✦Roadmap
How we think about AI here.
No committed dates, no marketing roadmap — just an honest list of what's real today and what's an idea for tomorrow.
Real today
Today
Adaptive streaming quality
Bitrate and resolution tuned to your network in real time, host-side, zero configuration. This is engineering, not AI.
Exploring
Exploring
Agents · predictive input · neural upscaling
Ideas we're exploring: agents driving a real desktop, predictive input, and on-device upscaling. Nothing ships yet, and we're not promising when.
Exploring
Exploring
Voice control · behavioural security
Ideas on our list: voice control and session anomaly detection. Nothing ships yet, and we're not promising when.
✦Zero-knowledge
If AI ever ships, it runs on your hardware — not ours.
No AI capability ships in Remio today. If and when one does, the design intent is the same as everywhere else: on-device or on your host machine. Remio's relay only ever forwards encrypted bytes it cannot decrypt — no Remio cloud endpoint touches your pixels or keystrokes.
What runs where
Adaptive streaming quality tunes on the host today — plain engineering. Any future AI capability, if and when it ships, follows the same rule: on-device or on your host, never through a Remio server.
No screen data for training
Remio does not collect screen content, keystrokes, or session data — for AI training or anything else, today or in the future. Any on-device model would run inside the client sandbox and never write frames to disk. The relay only ever sees opaque encrypted bytes.
✦Questions
Frequently asked questions
Will Remio's AI features run in the cloud?
No AI capability ships in Remio today. If and when agent or on-device AI features arrive, the design intent is the same as everywhere else in Remio: on-device or on your host, never a Remio cloud in the data path. Remio's relay only ever forwards DTLS-SRTP encrypted bytes it cannot decrypt — there is no Remio cloud endpoint that touches screen pixels or keystrokes.
Can I use Claude or GPT to control my Mac through Remio?
Not yet. Remio is built for people today. Its native architecture — direct input, direct screen access, a fast encrypted channel — is the kind of foundation AI agents will eventually need, but there is no agent platform shipping, and we're not promising a date.
Does Remio use AI to adjust streaming quality?
Remio already adapts bitrate and resolution to your network in real time, and that ships today — but it's engineering, not AI. Everything under the 'AI' banner on this page — an agent platform, on-device upscaling, predictive input, voice control — is exploratory. None of it ships yet, and no date is promised.
Will Remio use AI to predict my input or upscale video?
Maybe, eventually — these are exploratory ideas, not commitments. If they ship, the same rule applies as everywhere else in Remio: on-device, private, and no date promised until it's real.
Does Remio's AI collect screen data for training?
No. Remio does not collect screen content, keystrokes, or session data for AI training or any other purpose — today or in any future AI feature. If and when agent or on-device AI features ship, they will run on-device or on your host, never through a Remio cloud, and the relay will only ever see opaque encrypted bytes.
Adaptive streaming today. Everything else, honestly labeled.
Install Remio on the computer you want to reach and the device you reach it from. Adaptive streaming quality is on by default. Nothing else on this page ships yet, and we're not promising when it will.