WorkBuddy
Tencent's desktop workplace agent, the office sibling of CodeBuddy under Buddy AI, that operates the local computer to deliver documents, spreadsheets and presentations rather than just chat.
| Vendor | Tencent |
|---|---|
| Form | Desktop workplace agent (Win / Mac) |
| Launched | March 9, 2026 |
| ARscore | not scored |
| Pricing tier | Free (¥500 credits) / ¥99 / ¥199 / ¥999; ent ¥198/seat |
| Coding models | Hunyuan / DeepSeek / GLM / Kimi / MiniMax (selectable) |
| Open source | No |
Official links
Outbound, for trial and verification.
| Official site | https://www.workbuddy.cn/ |
|---|---|
| Pricing | https://www.workbuddy.cn/docs/workbuddy/Pricing |
| English interface / changelog | https://www.workbuddy.cn/ |
Capabilities
What it ships, in its own terms.
Reads and batches local files, generates documents and PPT, cleans and pivots Excel, runs web and software automation, and connects to Slack, Telegram, Discord, WeCom, Feishu, DingTalk, QQ and WeChat. Follows Anthropic Skills for reusable flows (e.g. one-tap typesetting to WeChat drafts) and MCP; high-risk command interception guards local execution. Shares credits with CodeBuddy.
Use cases
Scenario coverage (presence, not benchmark performance).
| Scenario | Local files | Docs / PPT | Excel reports | Analysis | Web search | Mini-programs | Web building |
|---|---|---|---|---|---|---|---|
| WorkBuddy | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ◐ |
✓ supported · ◐ partial or via a sibling product · ? unclear · — not in scope. Presence, not benchmark performance.
Users
Subscriber and active-user counts, source-labelled.
not published (a community 'blue book' hit 1k stars in two days) — Source: Community
Third-party review
Independent coverage, not vendor marketing.
Third-party editorial: no independent editorial score; Tencent's own WorkBuddy Bench ranks models, not the agent
Coding ability (benchmark): not scored — no DeepSWE or Terminal-Bench measurement.
No independent editorial score. Tencent publishes its own WorkBuddy Bench (Jul 24, 2026), a dual-harness ranking of models across Code/Web/Office/Security; it is vendor-run, not an independent measure of WorkBuddy itself.
Source: Vendor (WorkBuddy Bench)
Compare further
The reference agents and the scored Chinese entry.