WorkBuddy
Tencent (Tencent Cloud) · WorkBuddy: Tencent's Desktop AI Agent That Turns One-Sentence Requests Into Finished Reports, Decks and Spreadsheets
WorkBuddy is Tencent's desktop AI agent for office work on Mac and Windows: you describe a task in one sentence, and it plans the steps, works on files in folders you authorize, and returns a finished report, slide deck, spreadsheet or analysis. It suits office professionals and teams who want an agent that works on local files, can be sent tasks from WeChat, Slack, Telegram or Discord, and has a free tier with credit-based paid plans.
Best for
Office professionals, analysts and small teams who produce reports, decks and spreadsheets from files on their own computer, want an agent that can be tasked from WeChat, Slack, Telegram or Discord, and prefer a free tier with low-cost credit plans; especially teams already on Tencent tools such as WeCom, Tencent Docs or QQ Mail
Not ideal for
People who need agents that keep running with their computer off, organizations that require a documented data-processing region for the international edition before adopting, users on Windows 11 ARM64 or macOS older than 12, and buyers who want an open-source or self-hosted personal agent
Who it's for
Office professionals, analysts, operations and marketing staff, and teams who produce documents, decks, spreadsheets and research from local files, including enterprises on Tencent's WeCom and Tencent Docs ecosystem
Capabilities
- Assign work in one natural-language sentence; the agent breaks it into steps, runs tools and scripts, and checks its own progress until it delivers a usable file
- Reads and writes files in folders you authorize, with a separate workspace per task and a results panel for artifacts, workspace files, changes and previews
- Work Mode for documents, data analysis, slide decks, deep research, email drafting, weekly reports and batch renaming, sorting or converting files; Coding Mode for code generation, review and bug fixing
- Parallel tasks, plus automatically assembled agent teams for complex tasks (can be switched off in settings)
- Expert Center with domain-specific AI roles and expert teams in which a lead role splits work among members (Tencent Cloud cites 100+ preset experts)
- Skills from built-in, marketplace or uploaded packages, or generated from a description; uploaded skills get a security check, and high-risk ones always need manual confirmation
- MCP servers and connectors, including GitHub, Jira, Google Drive, Gmail and Outlook, plus Tencent Docs, QQ Mail, Tencent LearnShare, TAPD and Weiyun, and custom MCP connectors
- Scheduled automations that run a prompt once, daily, weekly, monthly or at hourly intervals, with run history and a test-run option
- Assistant remote control: send a task from WeChat, Slack, Telegram or Discord on your phone, have the desktop app run it, and get the result back in the same chat
- Default Permissions mode that stops for confirmation before high-risk operations outside the task's workspace, and a Full Access mode for trusted, recoverable tasks
- Local memory of preferences and context across sessions, with import from ChatGPT and Claude
- Auto model routing, built-in models by plan, and custom models through OpenAI-compatible Chat Completions APIs (DeepSeek, Kimi, Zhipu, MiniMax, OpenAI, Gemini, OpenRouter, Tencent Cloud Token Plan) or a local Ollama server
- Image and video generation, a built-in browser, Computer Use, a standalone Office window with in-document AI chat, and a Library for managing and sharing artifacts (per the changelog)
Limitations
- Scheduled automations and remote Assistant tasks run on your computer, so the docs require it to stay powered on, awake and running WorkBuddy, with a network connection; Tencent Cloud's product page mentions 24/7 cloud task hosting, but the international docs do not document how to use it
- Credit-metered: consumption depends on model tier, task complexity and context length, expert teams can use much more than a single expert, and the Free plan includes only 100 credits a month
- The list of built-in models is shown in the app rather than published, and the Free plan uses automatic routing instead of letting you choose every model
- Custom models must expose an OpenAI-compatible Chat Completions API, and the docs warn that a working connection does not mean tool calling, image input or long inputs will work
- The security FAQ says data is processed locally and that server-side fragments are discarded and never used for training, but the international docs reviewed do not state where cloud-side data is processed or stored
- Server or cloud deployment of the desktop agent is not yet supported, according to the FAQ; private and dedicated deployments are sold as enterprise editions through Tencent Cloud
- Windows 11 on ARM64 is not supported by the current version, according to the FAQ, and macOS 12 or later is required
- Pro monthly subscriptions and top-up packs are non-refundable; annual Pro is refundable only within 7 days and only if no plan credits were used
- Not open source, and no public API for the personal plans is documented; OpenAPI access is listed only for the Team plan
Use cases
- Pointing WorkBuddy at a folder of monthly sales exports and asking for an analysis report with charts and a summary deck
- Turning meeting minutes or a requirements document into a PPT, a weekly report or a structured Markdown summary
- Batch-organizing a Downloads or desktop folder: sorting images by date, renaming files and converting formats
- Scheduling a 9 a.m. daily briefing that searches the past day's industry news and saves a dated Markdown file to a workspace
- Sending a task from WeChat or Slack while away from your desk and getting the finished file back in the same chat
- Running a deep-research task with an expert team that splits analysis, drafting and review among several AI roles
Our take
WorkBuddy packs a lot into one desktop app for a low price: a free tier, a $10 Pro plan, and local-file execution, parallel tasks, skills, MCP, connectors, scheduled automations and phone-based remote tasking in one app. Its documentation is unusually candid about the limits. A successful model connection does not prove that tool calling works, a test run actually executes, and local models do not guarantee that data stays on the machine. The tradeoffs are structural. It is a desktop agent, so anything scheduled or sent remotely depends on a computer that stays awake. Credits are spent at different rates by different models and by expert teams, which makes heavy use harder to budget. Buyers outside China should also note that the international docs make strong security claims without naming a data-processing region. Individuals and small teams who build documents from local files will get the most from it, as will companies already on WeCom and Tencent Docs. Teams that need agents running in the cloud around the clock, or a documented data region, should evaluate the enterprise editions or a cloud-hosted alternative.
Who should use it
Knowledge workers who regularly turn local data and documents into reports, decks and spreadsheets and want an agent that does the file work for them; people who want to send tasks from WeChat, Slack, Telegram or Discord while away from their desk; cost-conscious users who want a free tier and a low-priced plan with model choice, including their own API keys or local Ollama models; and teams standardized on Tencent tools.
Who should skip it
Teams that need agents to run in the cloud while their computers are off, organizations whose compliance reviews require a documented data-processing region or an open-source codebase, users on Windows 11 ARM64 or older macOS versions, and anyone uncomfortable giving an AI agent write access to local folders.
Strengths
- Works directly on files in folders you authorize and returns finished, openable deliverables rather than chat answers
- Free plan to start and a low-priced Pro plan ($10 a month or $96 a year) with 2,000 monthly credits
- Remote tasking from WeChat, Slack, Telegram or Discord, with results returned to the same chat
- Broad extensibility: skills marketplace with security checks, MCP servers, connectors, expert roles and scheduled automations
- Bring-your-own models through OpenAI-compatible APIs or a local Ollama server, alongside built-in model routing
- Clear permission modes that pause before high-risk operations outside the task's workspace
Weaknesses
- Automations and remote tasks stop when your computer sleeps or shuts down
- Credit costs vary by model and task, so heavy use or expert teams can drain a plan quickly
- Built-in model list is not published, and Free users get automatic routing only
- No documented data-processing region for the international edition
- The agent runs as a desktop app, with no supported server or cloud deployment for personal plans
- Integrations run deepest in Tencent's own ecosystem
WorkBuddy pricing
Free
$0
- 100 credits per month
- Automatic model routing
- 3 automated tasks
- 250 one-time credits for new users, valid 14 days
Pro
$10/month or $96/year
Billed monthly
- 2,000 credits per month (1,000 base + 1,000 bonus)
- All built-in models
- 15 automated tasks
- First month $7 for new subscribers
- Early access to new features (rolling out gradually)
Team
$40/seat/month or $480/seat/year
Billed monthly
- 1,000 credits per seat per month in a shared team pool
- Admin console and unified billing
- Third-party SSO, organization management and per-member usage controls
- Enterprise skills, experts and model configuration, security and audit
- OpenAPI access
Credit top-ups
From $15 for 500 credits (Pro)
Billed one-time
- Pro pack: 500 credits for $15, valid 1 month
- Team packs: 2,000 credits for $40 up to 200,000 credits for $4,000, valid 6 months
- Non-refundable; requires an active Pro or Team plan
Free tier limits: 100 credits a month with automatic model routing and 3 automated tasks; new users also receive 250 credits on first sign-in, valid for 14 days.
Note: Prices are for the international edition at workbuddy.ai; the mainland China edition (workbuddy.cn) and Tencent Cloud's enterprise editions are priced separately. Credits are spent at different rates depending on model tier, task complexity and context length, and monthly plan credits do not roll over. During a limited-time promotion, active Free and Pro users earn 30 or 50 extra credits a day, Free and Pro users can create up to 99 automated tasks, and Free users can use all models; Tencent Cloud's enterprise page also lists 1,000 bonus credits per Team seat a month during a promotion. The pricing table also lists code-completion allowances, and the subscription terms count activity in either the WorkBuddy or CodeBuddy client, so plans appear to be shared with Tencent's CodeBuddy tools. Pro monthly plans and top-ups are non-refundable; annual Pro can be refunded within 7 days if no plan credits were used.
Technical specs
Modalities
Text, Image input, Documents, Spreadsheets, Presentations, Image generation, Video generation
Available models
Where WorkBuddy excels
Monthly reporting from local spreadsheets
An analyst can point a workspace at the month's exports and ask for a report with charts and a summary deck, getting files they can open and edit instead of instructions to follow.
Tasking the office computer from a phone
With Assistant bound to WeChat or Slack, a manager in a meeting can ask the desktop agent to draft minutes or prepare materials and receive the result in the same chat, as long as the computer stays awake.
Recurring briefings and weekly reports
Scheduled automations run a fully specified prompt daily or weekly against a workspace and keep a run history, which suits news briefings and Friday status summaries.
Teams on Tencent's office stack
Connectors for Tencent Docs, QQ Mail, TAPD and Weiyun and remote tasking over WeCom suit companies that already run on Tencent tools, and Team plans add SSO, organization management and per-member usage controls.
WorkBuddy vs. competitors
WorkBuddy vs. Claude Cowork
Claude Cowork is Anthropic's task-execution agent inside the Claude app on paid Claude plans, running Claude models across your files and connected tools. WorkBuddy is a standalone Tencent desktop app with a free tier and credit-based plans, routes across built-in models or ones you bring through OpenAI-compatible APIs or Ollama, and adds expert roles and remote tasking from WeChat, Slack, Telegram or Discord.
WorkBuddy vs. Manus
Manus runs tasks mainly in its own cloud sandbox, browsing, running code and returning deliverables, so its work does not depend on your computer staying awake; its desktop app can also work on authorized local folders. WorkBuddy is built around your Mac or Windows machine, processing local files directly, but its scheduled and remote tasks only run while that computer is on and running the app.
WorkBuddy vs. Perplexity Computer
Perplexity Computer is the agent in paid Perplexity plans, splitting goals across subagents and many frontier models with research at its core. WorkBuddy centers on producing documents, decks and spreadsheets from your local files, with a free tier, expert roles and skills, and model choice that includes bring-your-own and local models.
WorkBuddy vs. OpenClaw
OpenClaw is a free, open-source, self-hosted assistant that lives in messaging apps and appeals to technical users who want full control. WorkBuddy offers similar phone-based remote tasking through WeChat, Slack, Telegram and Discord, but as a closed-source commercial desktop app with a GUI, permission modes, a security-checked skills marketplace and, on Team plans, SSO and admin controls.
Frequently asked questions
What is Tencent WorkBuddy?
WorkBuddy is a desktop AI agent from Tencent Cloud for everyday office work. You describe a task in one sentence, and it plans the steps, works on files in folders you authorize, and delivers a finished result such as a report, slide deck, spreadsheet or analysis. It runs on Mac and Windows, and the same team also makes the CodeBuddy coding agent.
How much does WorkBuddy cost?
On the international site, the Free plan includes 100 credits a month with automatic model routing. Pro costs $10 a month (new users can get the first month for $7) or $96 a year and includes 2,000 credits a month (1,000 base plus 1,000 bonus) with all models. Team costs $40 per seat a month or $480 per seat a year, with 1,000 credits per seat in a shared pool and an admin console. Pro and Team users can buy one-time credit top-ups. Credits do not roll over, and usage varies by model, task complexity and context length.
Which AI models does WorkBuddy use?
WorkBuddy's built-in models are listed in the app's model selector rather than on a public page, and availability depends on your plan: Free uses automatic routing, and Pro unlocks all built-in models. You can also add custom models that expose an OpenAI-compatible Chat Completions API, with presets for DeepSeek, Kimi, Zhipu, MiniMax, OpenAI, Google Gemini, OpenRouter and Tencent Cloud's Token Plan, or run local models through Ollama.
Is WorkBuddy available outside China?
Yes. The international edition at workbuddy.ai has English documentation, US-dollar pricing and sign-in with Google, GitHub or X, and the app's interface supports English and several other languages. A separate mainland China edition is published at workbuddy.cn.
Does my computer need to stay on for WorkBuddy to run tasks?
For the desktop agent, yes. The docs say scheduled automations and tasks sent remotely through Assistant need the computer powered on, awake and running WorkBuddy, with a network connection, and they recommend enabling "Allow Running When Locked" for remote use. Tencent Cloud's product page also describes a cloud assistant with 24/7 task hosting that keeps working after the client is closed, and its enterprise lineup includes a separate Managed Agents service, but the international docs reviewed do not explain how to set up cloud-hosted tasks.
Does WorkBuddy upload my files?
WorkBuddy reads and writes files only in folders you authorize, and Default Permissions asks before high-risk operations outside the task's workspace. Tencent's security FAQ says data is processed locally, and that the server handles only data fragments, discards them after use and never uses them for model training. Tasks still send requests to the model you choose, and the docs note that even a local Ollama model does not guarantee data never leaves your computer when web search or connectors are used.
WorkBuddy vs Claude Cowork: what is the difference?
Both are agents for general office work that act on your files and connected tools. Claude Cowork is part of the Claude app on paid Claude plans and runs Anthropic's models. WorkBuddy is a standalone Tencent desktop app with a free tier and credit-based plans, routes across built-in models or models you bring, adds preset expert roles and expert teams, and can be sent tasks from WeChat, Slack, Telegram or Discord.
Integrations & fit
Alternatives to consider
Anthropic
Professionals with recurring document, spreadsheet, or research workflows who already pay for Claude and want them done autonomously
Paid
Manus
Professionals and teams who need AI to complete defined tasks end-to-end — research reports, data extraction, scheduled automation, presentations — rather than assist with each step manually
FreemiumPerplexity AI
Perplexity Pro, Max, and Enterprise users who want to delegate research-heavy, multi-step work, such as competitive analysis, financial models, recurring briefs, and decks, and who value cited sources and the option to keep sensitive files on their own hardware
Paid
OpenClaw Foundation
Technical users who want a private, self-hosted AI assistant they fully control, running through their own messaging apps and model of choice
FreeAbout WorkBuddy
WorkBuddy is built by the Tencent Cloud team behind the CodeBuddy coding agent, and it is the office-work counterpart to that product. You install the desktop app, sign in, pick or create a workspace (the folder a task reads from and saves to), and give an instruction such as "generate an analysis report with charts from this sales data" or "turn these meeting minutes into a PPT." The agent breaks the request into steps, runs tools and scripts, and puts its artifacts, file changes and previews in a results panel. Several tasks can run in parallel, each in its own workspace. Work Mode covers documents, data analysis, decks, deep research, email drafting and batch file processing, and Coding Mode covers code generation, review and bug fixing. Around that core sit the features that set it apart from a chat assistant. An Expert Center offers AI roles for specific domains and expert teams in which a lead role splits the work among members; Tencent Cloud's product page says there are more than 100 preset experts. A skills marketplace, uploadable skills that are security-checked before installation, MCP servers, and connectors extend what it can reach: the international docs name GitHub, Jira, Google Drive, Gmail and Outlook, and the connector page lists Tencent services such as Tencent Docs, QQ Mail, TAPD and Weiyun. Scheduled automations run prompts once, on a recurring schedule or at intervals. Assistant, the remote-control feature, lets you send a task from WeChat, Slack, Telegram or Discord on your phone and get the result back when the desktop app has finished it. Local memory carries preferences across sessions and can import memories from ChatGPT or Claude. Two permission modes govern risk: Default Permissions pauses for confirmation before high-risk operations outside the task's workspace, and Full Access is meant for trusted, recoverable work. Models are routed automatically on the Free plan, Pro unlocks the full built-in list (which is shown in the app rather than published), and you can add your own models through OpenAI-compatible APIs, including DeepSeek, Kimi, Zhipu, MiniMax, OpenAI, Gemini, OpenRouter, Tencent Cloud's Token Plan, or a local Ollama server. The tradeoffs come with that design. Everything runs on your computer, so the docs say scheduled tasks and remote Assistant requests only run while the machine is on, awake and running WorkBuddy. Usage is metered in credits that vary by model and task size, and expert teams can use far more credits than a single expert. The international edition (workbuddy.ai, priced in US dollars, with sign-in through Google, GitHub or X) is separate from the mainland China edition at workbuddy.cn. Tencent's own ecosystem, from WeCom and Tencent Docs to Tencent Meeting, is where its integrations run deepest.
Updates from WorkBuddy
Tencent launched Buddy One, a unified AI workspace that brings together Tencent Docs, Tencent Meeting and its cloud drive around the WorkBuddy agent, according to Caixin Global's report on the launch.
From August 7, 2026, Pro costs $10 a month or $96 a year and includes 2,000 credits a month (1,000 base plus 1,000 bonus). Pro subscribers who had auto-renewal on before that date keep the earlier $9.95 monthly price unless they cancel.
Are you the founder? Claim this listing →