WorkBuddy

WorkBuddy

Tencent's desktop AI work agent you can connect to TokenMix

What it is

WorkBuddy is Tencent's desktop AI work agent, part of the CodeBuddy family. You give it a task in plain language and it breaks the task into steps, works with your local files, calls external tools through MCP, and produces finished deliverables such as reports, spreadsheets, and slides. It ships with built-in models (Hunyuan, DeepSeek, GLM, Kimi, MiniMax) and a large set of expert roles, and, unlike most big-company AI apps, it lets you add your own custom model endpoint.

Before you start

Download and install WorkBuddy from its official site (workbuddy.ai, or codebuddy.cn/work in China). Then create a TokenMix account, open the console, and generate an API key, which looks like sk-tm-... Decide which model you want WorkBuddy to use — for example deepseek-v4-pro, qwen3.6-max-preview, glm-5.1, or kimi-k2.6.

Connecting WorkBuddy to TokenMix

In WorkBuddy, open Settings, then Model, and choose Custom — the option for a model that is not in the built-in provider list. Fill in three fields: the URL https://api.tokenmix.ai/v1/chat/completions, your sk-tm-... API key, and the model name you want, for example deepseek-v4-pro. If WorkBuddy reports a path problem, turn on the Custom Protocol toggle in advanced settings; that tells WorkBuddy to send requests to the exact URL you entered without rewriting the path. Save the model.

Verify it works

Select your new custom model in WorkBuddy, then give it a simple task, for example summarizing a document into three bullet points. If it runs and returns a result, the connection is working and the agent is now using your TokenMix model.

FAQ

WorkBuddy's custom models use the OpenAI chat-completions format, which is exactly what TokenMix's /v1 endpoint speaks, so any chat model TokenMix serves will work. This swaps the model the agent reasons with; WorkBuddy's own features, such as file operations, MCP tools, and expert roles, keep working as normal. The custom-model settings are stored locally on your own machine.