Another SSH client, 15 days, 663 stars: Motty didn't win on an empty niche, it won on the author's name and an engineering post

Motty is a native desktop SSH client written in Go: terminal, SFTP, port forwarding, remote desktop and an AI assistant in one window. The category is crowded. In the same month, seven or eight "native + AI" SSH clients appeared, and only this one reached a few hundred stars. The reason is simple: the author made DBX, a database client with 29k stars, the first post was well written, and he shipped 12 versions in 26 days. The playbook is worth copying; the niche itself isn't worth chasing.

The Oct 28 post

On the afternoon of Oct 2026 28, in linux.do's "domestic alternatives" section, user skylertong posted under the "open-source promotion" tag a thread titled "To make an SSH client faster and quieter, I first built a Go desktop framework: Motty × Viva." The first line:

There are already plenty of SSH clients, and I still wrote another one.

He then explained why he didn't use Electron or Rust's GPUI ("long compile times, big disk footprint"), so he wrote the desktop framework itself in Go. The framework, Viva, was open-sourced on GitHub the same day. Then came a set of numbers readers can check: on an M2 Pro, startup to usable took 80–103 ms, idle CPU stayed under 0.01%, installers were 6–10 MiB, and baseline memory was about 50 MiB. He also added a line saying he didn't want to use "native" alone as a performance claim.

The post got 520 likes and 255 replies. skylertong is t8y2 on GitHub, and DBX, built half a year earlier, now has 29,211 stars.

What it is

A remote workspace for three platforms (macOS, Windows, Linux) with no Electron and no WebView. The features listed in the post:

• SSH tabs, split panes, search, recording;
• Dual-pane SFTP file manager, resumable transfers, remote editing;
• Jump hosts, local/remote/SOCKS5 port forwarding;
• VNC / RDP remote desktop tunneled over SSH;
• One-click host import from OpenSSH, PuTTY, Xshell, MobaXterm, WindTerm, Electerm;
• An AI assistant with risk approval and redaction by default, plus an MCP interface for external agents such as Claude Code;
• Encrypted config sync via WebDAV, S3, and GitHub Gist.

Apache-2.0 licensed, free, with brew and scoop install sources, and the macOS build is signed and notarized.

How the stars grew

Data: GitHub daily star counts (10-29 to 11-06).

The curve has no spikes: a steady gain of 44 to 97 stars a day. Part of the reason is that the author kept posting new versions' changelogs in the original thread. From Oct 26 to Nov 6 he shipped 26 versions (v0.1.0 to v0.3.5), and the thread was pushed to the top again each time he posted v0.2.5 and v0.2.6 on pages 8 and 9. Installers for 26 versions have been downloaded about 4,500 times in total (GitHub, 11-06, excluding brew and scoop).

There are over 90 issues and PRs, mostly real requests written in Chinese, such as "MCP should support starting, incrementally querying and stopping long-running terminal tasks" and "sync often fails." There are 70 forks, and by Nov 5 someone had already posted on linux.sb a fork titled "Forking Motty, specialized for network engineering."

Almost nothing overseas: scattered reshares on Threads and Facebook, no big X accounts pushing it, and no Hacker News post. For now it is a project inside a Chinese developer community.

Why this one

The name leads. Neither V2EX nor the Telegram titles say what the product does; they only say "the DBX author." The hardest step of a cold start, getting people to click, was already done by DBX.

A good story with hard numbers. "To build one app well, I first built a framework" is a story on its own, and the performance numbers hit Chinese developers' dislike of Electron's "hundreds of MB, slow startup" (one linux.do commenter said exactly this about a similar app, Netcatty). The post also answered "why not Rust" up front, and the replies argued about it.

Zero migration cost. The import list covers Xshell, MobaXterm and WindTerm, the tools most used in China, so readers of the "domestic alternatives" section can switch at a glance.

Riding AI ops. Beyond the built-in AI assistant, it exposes MCP to external agents, turning an SSH client into an execution endpoint for Claude Code. Issues already include requests framed around that use case.

The two repos cross-link. People interested in the tool click Motty; people interested in "native desktop in Go" click Viva (now 200+ stars). Each repo points to the other.

Who it's competing with

On the closed-source side are Xshell, MobaXterm, FinalShell, and Termius, the users Motty is pulling from.

Oct 29, the second day after Motty launched, someone on linux.do posted asking, "Why are there so many open-source terminal projects?" The community is already tired of them. The clearest evidence is peer projects of the same age: shell-rs, also native plus AI, has 459 stars; Hatoba has 240; shellwing has 83. The feature lists are nearly identical. What differs is the author's name and that one post.

If it were you

Verdict: Copy the launch method, not the product. Don't build another SSH client. The category already has a veteran at 70k stars and seven or eight newcomers a month. Without a prior hit like DBX to build on, the same product will most likely stall under a few hundred stars (an inference, based on the peer projects).

Good fit for: Developers who already have a project with a few thousand stars and want the second one to ride its reputation; people who know linux.do and V2EX and will reply in threads every day and post changelogs; people who know Go or another niche desktop stack and can publish real benchmark numbers.

Not a fit for: First-time open-source authors with no prior project or community accounts; anyone who only wants one post and then to leave; anyone aiming at overseas markets (this playbook has only been proven in the Chinese-speaking community).

How to do it: Pick an old category people already use but complain is "too heavy in Electron" (database clients, Git clients, API debugging, log viewers, Redis/MQ admin...). Build a native, small, fast version, and:

• Write import support for every major competitor so switching costs nothing;
• Add an MCP interface so agents like Claude Code can call it, the newest angle for old categories;
• If you wrote an underlying framework or library along the way, open it as a separate repo and ship both on the same day so they boost each other.

How to get users: Launch on linux.do under the "open-source promotion" tag, with reproducible performance numbers and an explanation of "why not X" in the post. Each release, post a changelog in the original thread to keep it bumped for at least ten days. Two or three days later, post on V2EX with your earlier project's name in the title. Put community badges and a chat group at the top of the README.

When to do it: There's no rush, and this playbook doesn't depend on a trend. But the "add MCP to an old tool" angle was still fresh before the first half of 2027, and it gets more ordinary the longer you wait.

Signs you've misread this: ① Motty gets picked up by big overseas accounts (Hacker News, Reddit, X) and passes 3,000 stars within a month. That would show demand for "native + AI SSH clients" exists on its own, not just from the author's reputation, and the category still has room. ② Peer projects with no author halo, like shell-rs and Hatoba, also gain 2,000 stars in one or two months. Same conclusion. ③ Star growth stalls fast once Motty stops shipping releases. That means even the changelog tactic only works in bursts, so don't expect it to produce lasting traffic.

Data and sources: GitHub repo page and daily star counts (10-29 to 11-06), release download counts (11-06); linux.do thread /t/topic/3086528 (10-28) and "Why so many open-source terminal projects?" (10-29); V2EX /t/1253381 (10-31); reposts on zuphp.com, lvruan.com, ZNGG, and OKHK channels (11-02 to 11-03); linux.sb, "Motty fork, specialized for network engineering" (11-05); GitHub pages for Tabby, WindTerm, electerm, Netcatty, nyaterm, and shell-rs.