真实工作区A real workspace
选择本地目录,AI 直接读写文件、执行命令,不再复制粘贴。Pick a local folder. The agent edits files and runs commands directly — no copy-paste loop.
SmartWork 是运行在 Mac 上的 AI 工作搭子:在本地工作区读写文件、执行命令、改代码,多模型统一走 LLM Rank Hub 网关,每一步花多少 token 都算得清楚。 SmartWork is an AI teammate for your Mac. It reads and edits files, runs commands and ships changes inside your workspace — with every model routed through LLM Rank Hub and every token accounted for.
从读代码到改文件、跑命令、看 diff,一次任务从头做到尾。From reading code to editing files, running commands and reviewing the diff — one task, start to finish.
选择本地目录,AI 直接读写文件、执行命令,不再复制粘贴。Pick a local folder. The agent edits files and runs commands directly — no copy-paste loop.
同一套 OpenCode agent 运行时,工具调用、会话和权限都在本地。The OpenCode agent runtime, with tools, sessions and approvals kept on your machine.
DeepSeek、Claude、GPT、Gemini、Kimi、Qwen……统一走 LLM Rank Hub,按 token 计费。DeepSeek, Claude, GPT, Gemini, Kimi, Qwen and more — all through LLM Rank Hub, billed by tokens.
Default / Low / High 随时切换,支持推理的模型会把思考过程单独展示。Switch Default / Low / High at any time; reasoning models show their thinking separately.
⌘⇧X 直接进入命令模式,输出、退出码原位展示,不消耗模型 token。⌘⇧X switches to shell mode; output and exit codes render inline without spending tokens.
右侧直接看 token 计算器口径的用量、花费和 git diff,改动一目了然。Token-meter usage, spend and the git diff sit in the side panel — see exactly what changed.
目前开放 macOS 版。Windows 版正在打包与签名。macOS is available today. The Windows build is being packaged and signed.
首个版本先在 M 系列 Mac 上验证。Intel Mac 与 Windows 版会随后开放,内核与安装包都需要单独构建和测试。The first release is validated on Apple silicon. Intel Macs and Windows follow once their kernel sidecars and installers are built and tested.
所有模型调用统一走 LLM Rank Hub 网关,按 Token 计算器口径计量。你的工作区文件只在本机读取,只有对话内容会发给所选模型。Every model call goes through LLM Rank Hub and is metered by the token calculator. Workspace files stay local; only conversation content reaches the selected model.
可以用 LLM Rank Hub 账号的邮箱验证码、密码,或支付宝授权登录。勾选「30 天内免登录」后,本机会保存登录令牌,不再反复输入。Use your LLM Rank Hub email code, password, or Alipay. With “stay signed in for 30 days” checked, the token is kept locally so you are not asked again.
当前为手动更新:下载新版本后替换「应用程序」里的 SmartWork 即可,本地登录状态会保留。自动更新会在后续版本接入。Updates are manual for now: download the new build and replace SmartWork in Applications; your local sign-in stays. Auto-update comes later.
客户端免费。模型调用按实际 token 从 LLM Rank Hub 余额扣除,价格与网页版 Token 计算器一致。The app is free. Model calls are charged from your LLM Rank Hub balance by actual tokens, at the same rates as the website calculator.