Set up a local AI assistant team for your family and colleagues
Imagine a scenario where family members on weekends and team members on weekdays assign tasks to the same AI assistant from their respective messengers. Octop is an open-source platform that installs on your device to allow multiple users and AI agents to work together. You can interact with the same AI team via a web dashboard, command line, and various messenger apps.
A unified environment running in a single process
Octop handles the web dashboard, command-line tools, messenger channels, and scheduled tasks within a single process. All data is stored in the ~/.octop/ folder on your device and is not sent to external servers. An SQLite database is created upon first run, and you can switch to PostgreSQL if needed. This structure allows you to start the service directly on your computer or NAS without setting up a separate server.
Custom AI expert configurations per user
Each user can create their own team of 'Experts.' Each expert has independently configured workspaces, connected models, and available tools. You can apply one of 16 MBTI personality type templates to give the AI a unique character. Team members can share well-crafted expert settings or skills to reduce repetitive configuration work. The AgentTeams feature, currently in beta, helps coordinate multiple experts to perform complex multi-step tasks.
Interfaces extending from messengers to browsers
The chat interface is not limited to the web dashboard. You can connect various messenger platforms such as Feishu, DingTalk, WeChat, Telegram, and Discord to assign tasks to the AI in group chats. Browser automation features allow it to navigate web pages or take screenshots. Additionally, a remote desktop feature lets you view and control your computer screen in real-time from the dashboard. Support for the ACP (Agent Client Protocol) enables developers to invoke coding agents from an IDE or allows Octop to control external coding tools in a bidirectional integration.
Installation and initial setup
On macOS and Linux, you can run the installation script with a single-line command. On Windows, you download and run the installation file via PowerShell or cmd. The installation process uses uv to automatically configure a Python 3.12 environment in an isolated virtual environment, so Python does not need to be pre-installed on your system. After installation, you create the database and admin account with the octop init command and start the service with octop run. If you use Docker, you can proceed with the initial setup in the same manner.
Local-first design and security management
This tool operates on the principle of storing all chat logs, workspaces, and authentication credentials on your local device. JWT-based multi-user authentication ensures that each user's data is managed separately, and sensitive personal information is automatically masked before leaving the device. Dangerous operations, such as executing shell commands, are only performed after explicit user approval. You can freely switch LLM providers or storage backends, resulting in a structure that is not dependent on a specific vendor. However, since all data is stored locally, you must be aware that backup and security management are your responsibility.
