skales

The private AI that lives on your computer and does real work, not just chat.

📋 Table of Contents

• Demo
• ⚡ Why Skales?
• 🚀 What Skales Can Do
  • 🎯 Background Goals
  • 📐 Workflows
  • 🧩 Agent Skills
  • 🛠️ Skales Codework
  • 🏢 Organization
  • 👥 Teams
  • 📱 Skales Mobile
  • 🗣️ Voice: Talk to Skales
  • 🌐 Inline HTML Preview
  • 🖥️ Computer Use
  • 🦁 Lio AI (Code Builder)
  • 🎨 Skales Studio
  • 🗂️ Templates
  • 🌐 Built-in Browser Agent
  • 🌐 WordPress 2.0
  • 🎮 Playground
  • AIPointer ⦿
  • 🦎 Desktop Buddy
  • 📅 Planner, Calendar and AI Tasks
  • 🧠 Memory and Dreaming
• 🔌 Integrations
• 🧠 15+ AI Providers
• 🌍 Discover
• 📊 Skales Wrapped
• 🛡️ Privacy and Security
• 📦 Installation
• 🏗️ Architecture
• 🤝 Community
• ⭐ Star History
• 📜 License
• 🆓 Try Skales Free

Demo

⚡ Why Skales?

Skales is an AI agent that lives on your desktop. Not in a browser tab, not behind a restrictive API, not in a complex Docker container. It sits on your machine, has access to your files, your browser, your calendar, your email, and it does real work.

A 6-year-old built a game with it. A grandmother approved the setup.

🚀 What Skales Can Do

🎯 Background Goals

Type /goal and what you want, like /goal build me a trading bot, and Skales takes it on as ongoing work instead of a single reply. It plans the steps, keeps going on its own in the background with the chat closed, and picks back up where it left off after you reopen the app. Run more than one at a time, set one on a repeating schedule, or let Skales recognize a goal from how you ask. When it needs a decision it parks and shows a card to continue or stop; otherwise it keeps making progress. Every finished goal folds what it learned back into Memory, so the next one starts ahead.

💻 Code Mode (/code)

Point any chat at a folder on your computer and it works there like Claude Code, without leaving the conversation. A Chat / Code / Plan / Auto switch under the composer, on the New Chat screen (pick a folder and mode before your first message), or the /code command (/code <task> to start in one line) binds the folder: Plan investigates and proposes, read-only; Code edits directly; Auto runs the whole task on its own after a one-time consent. It patches files with edit_file instead of rewriting them, runs your build, installs and tests (test_run), and commits and pushes with your own git identity (git_commit / git_push, no added attribution). It renders git diffs inline in the chat as colored changes (green added, red removed) instead of plain text, and the model gets a repo map of the bound folder (which files exist and what each exports and imports) so it heads straight to the right file in a larger codebase. Shell commands get real time to finish (configurable up to 10 minutes), so installs, builds and deploys complete; it can deploy the folder with deploy_project (Firebase/Vercel/Netlify/npm) and ask you structured questions with a slide-up form when it needs a decision. A dedicated Chat & Code settings tab lets you give code work its own model (run a strong cloud model for code while chat stays on your default) and turn on deep reasoning (xhigh). If the folder is outside what Skales may touch, it asks first and widens access only the way you choose. Each chat keeps its own folder, so a normal chat is unchanged.

📐 Workflows

Workflows are the hand-drawn half of the goal system. A typed /goal lets Skales plan the steps for you; a Workflow lets you draw the steps yourself once, give them a trigger word like /goal-ship, and run that plan again whenever you need it. Think of it as a visual compiler onto the same plan format a typed /goal produces: you lay out steps, success criteria, and named inputs on a canvas, and Skales compiles them into the brief the planner already runs. One playbook store, four producers: the agent crystallizes finished goals into reusable plans, you draw the rest on a canvas, you teach a workflow by showing it once, walk Skales through a task in a normal chat, then turn that chat into a workflow and Skales distills the repeatable steps, trigger and success criteria for you, or you record a real desktop task: on the Workflow page hit Teach by recording, do it once on your screen (F10 stops, F9 pauses so you can skip a password), and Skales replays your exact clicks and typing when you run its /goal, with browser flows recorded in Playbooks instead. A learned workflow runs cold from then on through the normal goal path, falling back to vision when a button has moved. Workflows are opt-in, so you turn them on under Add-Ons, and a hint in Settings > Goal points you there. Running a saved workflow opens a fresh chat with its trigger prefilled and editable, so you can adjust the request before it starts.

🧩 Agent Skills (Open Standard)

Import skills from the Agent Skills format used by Claude Code, Codex, GitHub Copilot, and Cursor. Paste a GitHub URL, select a local folder, or paste SKILL.md content. Imported skills work across Chat, Codework, Browser, and Lio AI. Browse 1000+ community skills.

🛠️ Skales Codework

Select any project folder. Describe the task. Pick your model. Watch the agent read your files, plan an approach, write code, run