How big is WorkBuddy now?
PC monthly visits 8.85M (Mar) → 20.97M (Jun), ×2.4 in three months; IDC #1 in enterprise agents. Retention's odd too: active >60%, paid >80%.
I'm an AI & digital transformation consultant based in Malaysia. This site is just my running notes on WorkBuddy — not official, and nobody's paying me to write this. Three things I keep coming back to: what it's actually good for, how to stretch the free credits, and how it stacks up against ChatGPT / Claude.
WorkBuddy calls itself a "full-scene workplace AI workbench." In plain English: you throw it a task, and a whole team of expert agents behind it plans, does the work, and hands you the result — reports, decks, spreadsheets, code, research docs. Unlike chatbots that give you one answer and call it a day, this one feels more like a colleague actually working on your machine.
Tencent brought WorkBuddy into Malaysia in 2026 alongside its Johor cloud region, and that's when I started using it. The China version dropped Mar 9; the global version May 28.
All of these I tried on the free plan; didn't need Pro for daily work.
Give it an industry plus a question, and it finds sources, analyzes, writes the doc. I mostly just tweak the intro and outro.
Dump a messy transcript in, it pulls out the decisions and scope, and drafts the client proposal.
What you get is a finished thing you can use, not just an outline for you to fill in.
One prompt spins up a project, code goes straight to GitHub — no more begging someone to build a small tool.
Install a skill, gain a capability — like grabbing another expert for the team.
For real coding, hook up Tencent's CodeBuddy; hand it the plan to execute, same ecosystem.
There are roughly three flavors of agent products now. Figure out which camp WorkBuddy is in, and you'll know if it fits you.
| Dimension | Tencent WorkBuddy | OpenAI (ChatGPT agent) | Anthropic (Claude Code) |
|---|---|---|---|
| Where the work runs | Desktop app, reads/writes your own folders | Cloud sandbox on OpenAI's servers | Terminal / API / browser, dev-leaning |
| Models | Cross-vendor, auto-picked per task | OpenAI only | Claude only |
| Usable without coding? | Yes just talk to it | Chat product plus an agent mode | Assumes you know terminal/API |
| Free for daily use? | Yes includes daily activity bonus | Agent mode needs a paid tier | Billed per token, no daily free quota |
| How you extend it | Skill market / expert mode / MCP | GPTs and connectors | MCP (Model Context Protocol) |
💡 My own take: OpenAI's is clean but far from your files; Claude is powerful but leans dev. WorkBuddy is betting on the "local desktop" — an agent sitting on your machine, billed in free daily credits instead of per token.
Global pricing is honestly simple, and the free tier does real work — not a sample.
Desktop or browser, no card needed.
Lands automatically — covers research, docs, decks, code.
Open at least one chat daily; next day bonus credits show up — it basically feeds itself.
Sign up through my link, new users get a 350 bonus (check official terms).
100 credits/mo + daily bonus, +350 for new users.
Need more? Pro is $10/mo or $96/yr — 1,000 base + 1,000 bonus credits a month; there's also a 7-day Pro trial with 500 credits. Wait until the free tier genuinely runs dry before upgrading.
The facts and figures below are from public channels; the interpretation is all mine.
AI & Digital Transformation Consultant · Malaysia
I'm Dr. KG Lim, an AI & digital transformation consultant mostly running projects in Malaysia. WorkBuddy handles my research, docs, decks, and some of the code — freeing up time for the stuff that actually pays.
This site is just my own sharing: the mistakes I hit, the small tricks for stacking credits, the comparisons across vendors — dumped here, hopefully saving time for other Malaysia folks who want to use AI seriously.