Mindstone, a London-based AI transformation startup, launched Rebel this week—a local-first, agentic AI operating system designed to give enterprise teams predictable control over multiple AI models. The system, distributed under a "Fair Source" license, allows organizations with under 100 users to adopt and customize it freely, while larger teams need an enterprise license.
The marquee features include simplicity, extensive customizability around the open-source markdown standard, and an organizational memory layer. This memory ensures agents consistently use the enterprise's preferred AI model for each task or subtask, dynamically switching between local and cloud models to save costs and maintain data privacy and security. The system prioritizes visibility and predictability in how models are selected.
Rebel enters a crowded market of AI agent orchestration platforms, but differentiates itself through its local-first architecture and its Fair Source licensing model. By focusing on organizational memory, Mindstone aims to solve a core enterprise challenge: ensuring AI agents don't repeatedly need to relearn which model works best for specific tasks, reducing friction and improving reliability in knowledge-work settings.
For the enterprise AI space, Rebel signals a move toward systems that prioritize organizational consistency and data sovereignty over sheer model performance. The local-first approach appeals to industries with strict security or compliance requirements, while the fair-source licensing lowers the barrier for smaller teams to experiment with complex agent orchestration. The key trend to watch is whether organizational memory becomes a standard feature for enterprise AI platforms.
According to Greg Detre, Mindstone's chief technology officer, "Shared memory is the most empowering thing you could possibly do with a knowledge-worker AI." The company positions Rebel as a tool for teams with unique or specific workflows, rather than a one-size-fits-all solution.