Three accounts, 100 points signed in daily: how a WorkBuddy account-switching tool got to 1,750 stars in 82 days

workbuddy-switch is a desktop tool that helps people with several Tencent WorkBuddy / CodeBuddy accounts switch accounts with one click, auto check in, and use up the credits that expire soonest first. It never made GitHub's trending list and had no big-account shares. Stars grow steadily, about 20 to 30 a day. It rides on two things: Tencent's per-account credit subsidy, and posting two days ahead of every similar project. Anyone trying to build an open source project that gets stars can learn the playbook: when a platform launches a subsidy, be first to turn the hassle into a tool. But the WorkBuddy niche is already crowded, and Tencent is tightening up.

Aug 19, a post on linux.do

In 2026 Aug 19, a developer with the handle XBCHH posted "【开源工具】WorkBuddySwitch,workbuddy 切换工具" on linux.do. The opening wasn't about the product. It was about how he used it himself:

"I recently tried the points Tencent's codebuddy gives out to run deepseek. Speed and results seemed pretty good... so I signed up three accounts. With workbuddy, each account can also check in daily for 100 points."

He had Codex do the planning and CodeBuddy CLI run DeepSeek for execution. The annoyance was logging into each account daily to check in, so he wrote this small tool. At first it had only two features: account switching with auto check-in, and "automatic account rotation for CodeBuddy CLI, so points don't expire unused." The post got 123 likes. The repo was created three days earlier, on Aug 16. By Nov 6 it had 1,750 stars and 234 forks (GitHub, 11-06). It's the author's only public repo.

What it is

A desktop app written in Rust + Tauri under the MIT license, plus a web version on npm. It covers Tencent's AI coding family: WorkBuddy, CodeBuddy IDE, and CodeBuddy CLI, along with the CodeBuddy plugins for VS Code and JetBrains. Both the China and international versions are supported.

Every feature is for people who "have several accounts":

• Log in by QR code and switch the current account with one click;
• Auto check-in daily and keep tokens alive;
• Sort accounts by when their points expire and flag "use first"; the CLI follows this order to switch the default account automatically;
• Copy a session from one account to another, so context survives the switch;
• Points and token stats, plus a floating window that shows agent status.

It only changes the local login state file and offers no API. This matters later, because it's the big difference from similar tools with more stars.

Why people need it: points are per account, and they expire

The demand was created by Tencent's subsidy policy. WorkBuddy launched in March. By September it had 32.4M monthly active users, up 647% month over month (QuestMobile, reported by Sina and China Daily in October). To drive activity, it gives away points: new users get 250 points valid for two weeks, 100 per daily check-in, and 1,000 on the 7th consecutive day (Tencent Cloud Developer Community).

Points are issued per account, daily, and they expire. Each extra account is another free quota, but with many accounts three problems pile up: checking in on each one daily, tracking which account's points expire soon, and losing the conversation context after switching. workbuddy-switch handles all three.

Tencent's subscription terms say "no bulk registration or account sharing." The fact that they wrote this rule explicitly shows many people were doing it.

How the stars came: one post, and constant updates

Shipped first. It was created 8-16 and posted on linux.do 8-19. WorkDaddy, which has the most stars among similar projects, and workbuddy-auto-signin, the most popular check-in script, were both created 8-18, two days later. A reverse proxy that turned credits into an API appeared almost a month later in September. linux.do is where this demand first surfaced and concentrated. Later, WorkDaddy, WorkBuddy Manager, and various reverse-proxy tutorials all posted there, and the README's closing section specifically credits Linux.do.

Kept changing based on what users said. From v0.1.1 on 8-16 to v0.1.62 on 10-12, it shipped 56 releases and 421 commits in under two months. Randomized check-in times, skipping check-in per account, and enterprise-edition points were all added in response to issues. It also has 10 outside contributors. One of them, Mintimate, a blogger in the Tencent Cloud ecosystem, submitted PR #89. In issue #107, a user complained that after switching accounts in CodeBuddy IDE, "migrating chat content is hard." The author replied the same day and shipped a new version that day.

Focused on one platform, full coverage. The general-purpose switcher cockpit-tools has nearly 20k stars, but users in its issues #2327 and #2410 asked for CodeBuddy CLI and the international WorkBuddy version, and it still doesn't support them. Only workbuddy-switch covers these.

Then the long tail. Zhihu has a column about it, Bilibili has a video "WorkBuddy | Auto-claim points, one-click multi-account switching" (about 1,000 views), CSDN has a usage guide, and winget and trendshift list it. The npm version had 1,921 downloads in the last 30 days (npm, 10-07 to 11-05).

So the star curve is flat. 10-17 had 1,332 stars, 11-06 has 1,750, growing 15 to 30 a day. It gained +143 over the last 7 days and +131 over the first 7 days. The biggest single-day gain in the observation window was +52 (10-14). No spike, and no sign of star farming. The fork ratio is 13%, higher than typical repos, which suggests many people actually pull it down and use it.

Real install base has to be estimated from installer download counts. The latest Windows installer has about 15k downloads, the dmg about 1,600, and the Linux package about 800 (GitHub, 11-06). Users are probably somewhere between a few thousand and 10–20 thousand machines (estimate). The 570k cumulative downloads are mostly auto-update pulls of version files, so they shouldn't be counted as installs.

Who it's competing with

This is a crowded gray-market niche. Searching GitHub for workbuddy turns up over 8,000 repos. The same demand has more than a dozen projects with 100+ stars (star counts as of 11-06):

You can see that within the same wave of demand, the reverse proxies have the most stars. They move points out of WorkBuddy for use elsewhere. workbuddy-switch deliberately stayed out of that lane. In the "local account switching" slot it was earliest and most complete, but across the whole ecosystem it isn't the largest. Someone on linux.do said "it's not a reverse proxy, so it's safe," and that lower bar made low-tech users more willing to try it. This helped it hold its position, but the initial boost still came from being first and from the subsidy.

Where it stands now

Tencent is tightening up. 10-17 and 10-23 each saw a large wave of account bans. In issue #189 someone wrote "6 accounts dropped, 4 of them, and the one I'm using got dropped too." Under #198 there's a string of "same here." Someone also asked "where did everyone register their accounts? I want to register a new batch." Others worry too: "One has already stopped, I don't know if this one will keep updating."

The author's last push was v0.1.62 on 10-12. Since then there's been over three weeks of silence, and the reason isn't visible from outside. Stars are still rising, but that's running on existing traffic. The reverse-proxy project workbuddy2api-hub, in the same niche, announced it was no longer being updated on 11-06. How long tools like these last depends on how Tencent manages them.

If it were you

One-line verdict: Don't build another WorkBuddy switcher. This slot already has a first mover, the niche is crowded, and the platform is cracking down. What's worth learning is the playbook: when a big company's AI tool first hands out expiring per-account credits, whoever makes multi-account management smooth first can pick up a steady stream of stars in the community.

Who it suits: Developers who already hang out in communities like linux.do, use the product heavily themselves, and can ship several versions a week in response to issues. You're going for a few hundred to a thousand-plus stars and a name in the circle, not a project you'll maintain for years.

Who it doesn't suit: Anyone who wants a long-term project to put on a résumé as a flagship piece; anyone uncomfortable with their project being tied to "bulk registration, violating platform terms"; anyone unwilling to ship a same-day release every time the platform changes.

How to do it:

• Watch AI coding tools like Trae, Qoder, and Kiro that are handing out credits. Check whether their credits are per account, issued daily, and expiring. If all three hold, there's demand for multi-account management tools.
• Port the design directly: expiry-first ordering, automatic CLI rotation, auto check-in, and carrying the session across switches. Carrying context across accounts is the part few people build and users want most.
• Go deep on one platform and cover all its surfaces (desktop, CLI, IDE plugins, China and international versions). General-purpose all-in-one switchers always have gaps, and that's where your position is.
• Local changes only, no reverse proxy. You'll get fewer stars, but the project is less likely to be named and banned by the platform.

How to get users: Post an open source promo on linux.do, opening the way XBCHH did, explaining how many accounts you run, how you use them daily, and which step annoys you. Then reply on help threads like "points aren't enough" or "what do I do after getting banned." Add a Zhihu or CSDN usage guide, and get listed on winget.

When to be early: Within the first week after the platform announces its credit giveaway. workbuddy-switch was only two days ahead of the runner-up, but two days was enough to make it the first one recommended in the switching slot.

Signs you've read it wrong: More than a dozen similar projects appear on GitHub within three days of a new platform's subsidy launch, or the platform switches to per-person credits in its first month, or it starts mass-banning multi-account users. The first means too many people jumped early. The second means the demand is gone.