rohit267/net-merge

The problem

Most machines sit on one uplink at a time: Wi-Fi, or Ethernet, or a phone tether. When a large download, a game update, or a batch of cloud uploads matters, the other links stay idle while the active one saturates. net-merge treats all adapters as one pool, opening successive flows across whichever links are available, so a workload made of many connections spreads out instead of queueing behind a single pipe.

What's different

The usual answers are either hardware (bonding at the router or with a multi-WAN device) or a commercial aggregator, and both cost money or ISP cooperation. net-merge is a local TUN-based daemon that needs no account, no server, and no proxy configuration, and it claims to handle every app's TCP and UDP traffic including IPv6. The two capabilities that actually distinguish it are per-flow scheduling with a preference for faster links, and a fail-safe design where a crash leaves normal connectivity intact. Its own README is honest about the boundary: a single connection never gets faster, only parallel flows do.

Who it's for

People who regularly have two or more usable links and workloads that open many connections: Steam and game downloads, package managers, torrent clients, download managers, browser multi-connection downloads. It fits laptops moving between office Ethernet and a hotspot, and users on metered mobile data who want to combine a tether with home Wi-Fi. It is not for anyone expecting a single large transfer to accelerate, nor for readers who need per-app routing, fallback rules, or per-app bandwidth caps — the README points those users to the author's paid RoutEngine instead. Server operators wanting true link bonding should stay with router-level or ISP-level solutions.

Cost of adoption

Installation is one curl or PowerShell line, or an OS package; the README also documents apt, dnf, zypper, pacman, apk, and Homebrew paths. The project asks for elevated network rights: setcap on Linux packages and a sudoers rule on macOS so only the tray binary can run as root. Linux tray needs a StatusNotifierItem-capable desktop, and firewalls may require trusting the net-merge interface. The license is MIT. Windows binaries are unsigned, so SmartScreen and antivirus warnings are expected. The README is unusually detailed on install, but it is still a young project — no long-term operating history exists to judge stability.

Verdict

This is a 12-day-old repository that gained roughly 174 stars per day and briefly reached #2 on a Go trending board, with no Hacker News discussion and no daily-board appearance — the profile of an early spike rather than an established tool. The 2,125 stars and 112 forks suggest genuine interest, but the project has no track record yet. The concept is well understood and the engineering risks (kernel TUN, elevated privileges, multi-platform packaging) are real. Try it on a spare machine or one workload you can afford to lose before making it part of your daily setup.