# USBridge Remote is the software half of a hardware remote control line > USBridge Remote pairs a client application with an agent that captures the screen and injects input on the target machine, over Tailscale peer to peer networking, with a browser client and native builds for six platforms. The open tier is built on the Sunshine streamer, and the repository also carries the firmware and release tooling for the vendor's hardware line. **USBridge-Technologies/USBridge-Remote** — Software-only version of the USBridge ecosystem for secure, low-latency remote desktop control and peer-to-peer networking. - Repository: https://github.com/USBridge-Technologies/USBridge-Remote - Stars: 656 · Forks: 40 - Language: Go - License: GPL-3.0 - Published: 2026-10-21 · Updated: 2026-10-21 · Language: en - Canonical page: https://hysenlabs.com/projects/usbridge-technologies-usbridge-remote ## One interface for a hardware path and a software path USBridge Remote is described as a unified high performance client for managing remote machines, engineered to combine hardware level BIOS access through USBridge KVM devices with software based remote desktop in a single interface. That combination is the product. A KVM over IP appliance gives you a machine before it boots, which is the only way to reach a server whose operating system has crashed, whose disk is encrypted, or whose bootloader you are about to change. Software remote desktop gives you a session once the machine is running, with a frame rate and a control latency that make it usable for work rather than only for administration. The repository here is the software only version of that ecosystem, and the description in the repository listing says so, calling it a software only version for secure, low latency remote desktop control and peer to peer networking. The hardware integration section makes the relationship explicit, describing native integration with the USBridge KVM 2.0 appliance for out of band, bare metal management before OS boot, and the repository even carries an esp32-acm directory, which is where the firmware for the vendor's own controller board lives. ## A client on your desk, an agent on the target The architecture is the familiar split, and the README describes both halves. The agent runs on the target machine, the server or PC you want to reach, and it handles screen capture, input injection and Tailscale networking. The client is the control interface, installed on a workstation or laptop or run in a browser, and it manages connections, live remote desktop, virtual device passthrough and a snapshot registry. That last item is worth pausing on, since a snapshot registry in a remote desktop client means the tool can capture and restore the state of the remote machine, which is the software equivalent of what a KVM does with a physical session. Tailscale is doing the networking work, and that choice has consequences. Because the transport is a peer to peer mesh rather than a relay, there is no third party server in the path, and the agent shows a Tailscale address that you connect to directly. For a machine on a network you do not control, that is the difference between a working remote session and a request to open a port on a router. ## Six platforms and a browser client that needs no install The download matrix is wide, and it is organised by architecture rather than by operating system alone. For x86_64 there is a Windows zip and a Linux AppImage, and the web client opens in a browser. For ARM64 there is a macOS disk image, a Google Play app for Android and an App Store app for iOS, plus the same web client. The agent side is narrower, with Windows and Linux for x86_64 and macOS for Apple silicon, and no Android or iOS agent, which makes sense since the agent is the thing that captures a screen. A self updating APK is published on the latest release for people who would rather not use a Play Store account, and the mobile apps are separate from the desktop builds. The zero install web client deserves a caveat from the README itself, which notes that it operates with some feature and performance limitations because of browser security sandbox and WebRTC constraints, and recommends the native applications for the full experience. ## The web client waits for a certificate to exist The way the browser client reaches an agent is the most interesting technical detail in the README, and it is a security design rather than a convenience. The web client reaches an agent through its trusted label and device name, in the form of a hostname ending in device.usbridge.io. The certificate for that name is obtained the first time a browser comes to the agent, and it is only issued while you are signed in to your USBridge account in the agent, which means the very first attempt can take a few minutes. The agent shows a status row with the certificate so you can see the state, and the agent documentation covers the platform notes. The reasoning is that a browser will not talk to an agent on a bare IP address without a trusted certificate, so the agent has to be able to obtain one first, and that step is gated on an authenticated session. A KVM device also serves the web client itself on a port and moves to its own trusted name when it can. For an administrator, the practical takeaway is that the first connection to a new agent is slow, and that this is expected behaviour rather than a misconfiguration. ## Three tiers, and the open one is Sunshine The feature table is where the commercial structure shows, and it is worth reading carefully rather than skimming. The first column is labelled basic functionality and is marked as Sunshine open source, standard game streaming, with a built in open source Sunshine streamer, low latency desktop and gaming access, a bi-directional file and text clipboard, multi monitor display switching, integrated Tailscale peer to peer networking, and native Wayland support with promptless capture. The second column adds the free tier of the vendor product, described as a Rust based stream engine, with an instant connect custom remote protocol, raw USB peripheral device passthrough, adaptive streaming optimised for Wi-Fi stability, HDR10, headless virtual display management, low latency gamepad controller passthrough, browser client access with no installation, and Windows pre-login access for secure credential entry. The third column adds the professional tier, with lossless 4:4:4 chroma colour accuracy and Wacom tablet support with pressure and tilt. The two columns of extras sit under headings that distinguish the free tier from the professional one, so which of those capabilities you may actually use depends on the licence attached to your account, and the README does not spell that out in prose. ## Getting connected takes three steps The quick start is short and the order matters. You install the agent on the machine you want to reach and launch it, at which point it displays a connection token and a Tailscale address, and you connect Tailscale if you need access over the internet. Then you install the client on your workstation, laptop or phone. The third step is to add a connection by entering the IP or Tailscale address shown in the agent window. There is no account step in this sequence beyond the sign in already required for the certificate flow, and no configuration file to edit, which makes it one of the shorter remote access setups. The hardware story extends the same idea. The KVM appliance gives out of band access before the operating system starts, so a machine that will not boot is still reachable, and the pre-login access feature in the software tier addresses the same problem from the other direction by letting you type credentials into a secure screen on the Windows logon prompt. ## A vendor repository with firmware and release tooling included The tree tells you what kind of project this is. There are separate agent and client directories, a docs directory holding the multi language documentation, including English, German, French, Italian, Spanish, Brazilian Portuguese, Ukrainian, Polish, Japanese, Korean and Chinese, an assets directory with the diagrams, a scripts directory, a fastlane directory for mobile release automation, the esp32-acm directory for the controller firmware, and a .claude directory for agent configuration. The release naming is not semantic versioning in the usual sense: the recent tags are release-v3.1.19, release-v3.1.20 and release-v3.1.21, published on consecutive days in October 2026, which says that this ships often and that pinning a version matters more than usual in a rolling release line. The licence is GPL-3.0, the last push was on 2026-10-20, and the project is not archived. The interesting open question for a reader is the boundary between the GPL parts, which include the agent and the client, and the commercial capabilities that the hardware and the Pro features depend on. ## Conclusion USBridge Remote suits someone who needs out of band access to a machine, particularly one they have to reach before the operating system is up, since the product is built around a hardware BIOS control path as well as the software desktop path. It is a poor choice if you need every listed capability on day one, because the feature table splits across a Sunshine based tier, a free tier with a Rust stream engine and USB passthrough, and a professional tier with lossless chroma and tablet support, so the licence of each feature has to be checked. Verify three things before you deploy it: which tier you are entitled to use, whether the agent can run on your target platform, since the agent matrix covers Windows, macOS and Linux for x86_64 and Apple silicon only, and whether your network allows the Tailscale path, since the agent displays a Tailscale address and you connect Tailscale for access over the internet. The licence is GPL-3.0, releases are named release-v3.1.x, and the last push was on 2026-10-20. ## FAQ ### What is the difference between the USBridge Remote client and the agent? The agent runs on the target machine and handles screen capture, input injection and Tailscale networking. The client is the control interface on your workstation, laptop or phone, and it manages connections, live remote desktop, virtual device passthrough and a snapshot registry. ### Which platforms can the USBridge Remote client run on? For x86_64 there is a Windows zip and a Linux AppImage, and for ARM64 a macOS disk image plus Android and iOS apps, along with a zero install web client at web.usbridge.io. The agent itself is provided for Windows and Linux on x86_64 and macOS on Apple silicon. ### Why does the first connection to a USBridge Remote agent take minutes? The web client reaches the agent through a trusted hostname, and the agent obtains the certificate for it the first time a browser comes to it, which only happens while you are signed in to your USBridge account in the agent. The agent shows the certificate state in its status row. ### What open source software does USBridge Remote build on? The basic functionality tier is built on the Sunshine open source streamer, providing low latency desktop and gaming access, clipboard sharing, multi monitor switching, integrated Tailscale networking and native Wayland capture. Higher tiers add a Rust based stream engine, USB and gamepad passthrough, HDR10, pre-login access, lossless 4:4:4 chroma and Wacom tablet support. ## Sources - [Issues](https://github.com/USBridge-Technologies/USBridge-Remote/issues) - [License: GPL-3.0](https://github.com/USBridge-Technologies/USBridge-Remote/blob/main/LICENSE) - [README](https://github.com/USBridge-Technologies/USBridge-Remote/blob/main/README.md) - [Releases](https://github.com/USBridge-Technologies/USBridge-Remote/releases) - [USBridge-Technologies/USBridge-Remote on GitHub](https://github.com/USBridge-Technologies/USBridge-Remote) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/usbridge-technologies-usbridge-remote