LobsterAI

LobsterAI is an open-source desktop app that functions as a desktop-grade AI agent. It is designed to automate a wide range of office tasks such as data analysis, slide deck creation, document writing, and web research by interacting directly with the local operating system. The application is available for macOS, including both Intel and Apple Silicon architectures, as well as Windows.

Unlike standard chatbots that operate in isolated sandboxes, this software utilizes a Cowork mode that connects to local files, the terminal, and the browser. This allows the agent to execute commands and run tools within the user's actual working environment to perform real-world actions. Users can also control the desktop application remotely from their phones via various messaging platforms, including Telegram, Discord, WeChat, and Feishu.

Key features

• Direct access to local files, terminals, and browsers for task execution
• Integration with Model Context Protocol (MCP) servers for external data
• Remote command capability via WeChat, Telegram, Discord, and Feishu
• Creation of custom agents with specific identities and skill sets
• Support for scheduled recurring jobs and automated daily digests
• File-based persistent memory to track user preferences across sessions
• Built-in Python runtime for Windows to handle skill dependencies
• Ability to hot-load new skills and Expert Kits at runtime

The application is built on the OpenClaw agent framework and developed using Electron, React, and TypeScript. It utilizes a local-first architecture where sessions, configuration, and memory remain on the user's device. To maintain security and transparency, every tool call and sensitive action is gated behind user approval and logged for auditing. The system includes 28 built-in skills and allows for the creation of new ones via a dedicated skill-creator tool.

LobsterAI is a productivity tool for users who need an autonomous assistant capable of performing complex, multi-step workflows across different desktop applications and local data sources.