Tencent's Buddy One Makes a Single Prompt Reach Across Docs, Meetings, and Cloud Drive
Tencent's Buddy One is a concrete attempt to break the data silos that limit cross-app AI agents, and its bundled subscription model mirrors what Google and Microsoft already do with Gemini and Copilot.
Buddy One is Tencent's cross-app AI collaboration system that links WorkBuddy, Tencent Meeting, Tencent Docs, ima, and Tencent Cloud Drive. After linking a free Buddy One ID, WorkBuddy can retrieve meeting minutes from Tencent Meeting, structure them into a Tencent Doc, store a summary in ima, and archive raw files to Cloud Drive based on a single natural-language instruction. Credits from the separate apps merge into one pool, so a task doesn't stall when one app's quota runs out. The author demonstrates six workflows: turning meeting minutes into an archive, querying to-dos, generating a weekly report from multiple meetings, searching for GEO-related materials across apps, drafting an article from knowledge base entries, and backing up a local project folder with a generated material description.
Buddy One VIP costs 700 yuan per month (marked from 1,399 yuan, limited-time 50% off), which is 126 yuan less than buying WorkBuddy flagship, Tencent Meeting Pro, and Tencent Docs Super Member separately. VIP includes 50,000 universal credits monthly and 1TB shared storage across the apps. Existing WorkBuddy flagship subscribers can upgrade to VIP for free. The author argues that the real advantage is contextual: when meeting, document, knowledge base, and file data live in separate apps, an agent can only see one silo. Connecting them gives AI a larger context window, which is why Tencent is building this unified layer.
Data integration is the real bottleneck for cross-app agents: a single prompt across four apps only works when accounts, credits, and file access are unified, which is exactly what Buddy One does.
Tencent is playing catch-up to Google and Microsoft by bundling AI subscriptions, but its advantage is owning the entire office productivity stack, so it can control the integration layer directly.
The plan to add memory interoperability in phase two suggests that the first phase is purely about data plumbing; agents become genuinely context-aware only after user memory is shared.