
The footer lists an open-source license
Sending documents, spreadsheets, and websites to online tools often means the files leave the machine. WorkAny keeps those outputs inside one desktop agent and says the data stays on the computer. Folder cleanup looks at file type, content, and context, not only the extension. A described website becomes a styled front-end app with a live preview you can revise while watching it. Spreadsheets can include formulas, and slide decks cover structure, copy, and layout. Code runs in an isolated sandbox provided by Codex CLI. Skills can be written and shared, and external tools connect through the Model Context Protocol. It fits people who want documents, spreadsheets, and page prototypes produced locally, and developers willing to configure their own model provider.
WorkAny is a desktop agent that runs on your own computer and carries out tasks from natural language. It can organize folders and generate slides, spreadsheets with formulas, documents, code, and websites. The site says this work happens on the machine and that data stays there. Code runs in an isolated sandbox, and artifacts can be previewed live. Besides built-in models, it can use several model providers, custom skills, and connections to external tools. The footer lists an open-source license. The installer is on the website, and the source is on the code host linked from the page.
Difficulty: Intermediate
Organize folders by content
It organizes a messy folder using file type, content, and context, producing a clearer structure rather than sorting by extension alone.
Generate a local site preview
A short description becomes a React application styled with Tailwind CSS, with a live preview on the machine that can still be edited.
Documents, sheets, and slides
It can write and export reports, build spreadsheets with formulas, and arrange a slide deck's structure, copy, and layout from an idea.
Sandbox runs and custom skills
Code runs in an isolated sandbox provided by Codex CLI. Users can create and share agent skills, and connect external tools through the Model Context Protocol.
Tidy a local downloads folder
An operations or engineering teammate hands a mixed downloads folder to the desktop agent and has it regroup files by type and content.
Build event slides from a brief
A marketer writes the event points in natural language and has the agent build slide structure, copy, and layout on the local machine.
Draft a page with live preview
A product manager describes a page, gets a previewable front-end page locally, and keeps editing from that preview.
The site describes it as a desktop agent. Slides, spreadsheets, documents, code, and websites are generated on your computer, and the page says data stays there. Code runs in an isolated sandbox provided by Codex CLI, rather than by uploading files to an online editor. The installer and the source link are both on the same page.
Yes. The page includes built-in models and says a model is matched to the task. It also supports multiple providers, naming OpenRouter, Anthropic, OpenAI, and custom providers. The first start step is to configure a model, choose a built-in model or your own provider, then connect and deploy. Skills and external tools are separate: you create skills yourself, and external tools connect through the Model Context Protocol. Install does not come with a fixed tool list already set up.
The page lists five kinds of work: organizing folders, generating a previewable website from a description, writing and exporting reports, building spreadsheets with formulas, and arranging slide structure, copy, and layout. Each starts from natural language and can be split into a multi-step flow. External tools can be connected, and custom skills can be created and shared.
Real reviews and feedback from users