Local-first desktop AI agent for everyday automation
skales, built by Skalesapp, is an open-source, local-first desktop AI agent that turns advanced agent workflows into a native application for everyday users. The app runs as a standard native installer across Windows, macOS, and Linux, hosting autonomous agents that set high-level goals, manage email and calendars, perform browser tasks, and aid coding and research. Its design emphasizes minimal setup, on-device data handling, and extensibility through the Model Context Protocol, targeting non-technical users, privacy-conscious individuals, and developers who need an extensible agent host.
Handles autonomous goals and routine desktop workflows
skales implements an autonomous task execution model that runs agents against user-defined goals, and offers native integrations such as Gmail, Telegram, and system calendars. Browser actions run through Playwright, while SKILL.md teaches portable text-based agent skills. Typical outcomes include:
- automated email triage and scheduling
- web data extraction
- code scaffolding and review assistance
Agent outputs depend on the selected model and need verification
The app supports multiple providers including OpenAI, Anthropic, Google, OpenRouter, DeepSeek, and local Ollama instances, so generated results vary with the chosen model and API key. Skales uses a bring-your-own-key approach for cloud models, which means factual or high-stakes responses require independent checking. Autonomous agents can complete multi-step tasks, yet their reliability tracks the strengths and limits of the external model in use.
Installs natively but has platform and runtime expectations
skales ships as a native application (.exe, .dmg, .AppImage) and can act as an MCP-compatible host or client on local networks, with Android access and a web interface available. The app runs with roughly 300 MB idle RAM, making background operation feasible on typical desktop hardware. Some technical users report interface quirks tied to its Electron-based architecture, which may affect integration on specific systems.
Low-friction setup and on-device data handling for privacy-conscious users
The developer designed the app for zero-configuration installs and a visible desktop companion to reduce intimidation for non-technical users. Conversation histories, memories, and API keys remain on-device (for example in ~/.skales-data), and requests to cloud providers route directly from the user's machine. The Model Context Protocol support lets the tool join existing workflows without transferring core data off the desktop.
Practical choice for local automation with a model-dependent accuracy caveat
skales is a pragmatic option for non-technical users and developers who need an on-device agent to reduce repetitive desktop work. Generated outputs depend on the external model selected, so critical decisions require human verification. Treat the tool as an automation host that accelerates routine tasks while preserving local control, and run small experiments before delegating important workflows.
