Vellum is an open-source personal AI assistant platform designed to function as a persistent, autonomous digital companion across desktop and messaging environments. Unlike traditional chatbot interfaces, the project focuses on long-term memory, identity, proactive behavior, and real-world tool usage, enabling assistants to evolve alongside the user over time. The system integrates with macOS, Telegram, Slack, SMS, and additional communication channels while maintaining shared memory and context across platforms. Its architecture combines local-first storage, tool orchestration, sandboxed execution, and extensible workflow automation to allow assistants to read files, manage schedules, send messages, browse the web, and control applications. The repository includes a Bun-based runtime, CLI tooling, gateway services, and native macOS clients designed for modular deployment and customization.
Features
- Persistent cross-platform AI assistant identity
- Long-term memory and preference retention
- Integration with macOS, Telegram, Slack, and SMS
- Tool execution and workflow automation capabilities
- Local-first storage and sandboxed execution
- CLI, native desktop client, and gateway architecture