London continues to be a hub for enterprise AI innovation. Mindstone, a UK-based AI transformation startup, has launched Rebel, an agentic AI operating system designed to be local-first and distributed under a Fair Source license. The platform addresses a critical challenge for companies: orchestrating multiple language models transparently, securely, and cost-effectively without losing data control.
A local-first architecture built on Markdown files
Rebel stands out for its local-first architecture. Unlike frameworks like LangGraph or CrewAI that require complex databases and cloud infrastructure, Rebel stores agent state, prompts, and instructions in plain Markdown text files. These lightweight files can be inspected, modified, and moved easily. The primary configuration file, agents.md, serves as the central instruction layer and runtime boundary. This choice reduces API costs because formats like Word or PDF consume tokens for formatting, while Markdown keeps text clean. It also avoids vendor lock-in: if a company decides to switch platforms, all automations and memories remain accessible as local files.
Sponsored Protocol
Shared memory for collective intelligence
Greg Detre, CTO of Mindstone, explained that shared memory is the most empowering feature for an AI assistant. Rebel uses a tiered memory structure: high-value information is written to project-specific readme.md files; medium-value data becomes reference links to historical records; low-priority material remains in an indexed directory, ready to be recalled when needed. This approach avoids the common pitfall of dumping massive amounts of data into a database and hoping search will retrieve the right context.
Multi-model orchestration for flexibility and savings
One of Rebel's key features is its ability to break a task into subtasks and route each to the most appropriate model. A powerful cloud LLM handles complex planning; a cheaper model takes care of routine work; a local model manages sensitive steps, ensuring critical data never leaves the device. This is crucial for companies that must comply with privacy regulations or want to control costs. Detre noted that a user can simply ask for help, and the system automatically knows what is personal, sensitive, or shareable with the whole company.
Sponsored Protocol
Skills, operators, and automations for repeatable workflows
Rebel lets users create Skills, reusable multi-step procedures, and Operators that adjust agent behavior based on context, such as reviewing a pitch deck from an investor's perspective. Automations run scheduled background tasks, like scanning messages or drafting responses before an employee opens the app. These tools make AI not just reactive but proactive, embedded in daily workflows.
Sponsored Protocol
ROI dashboard to measure business impact
For larger organizations, Mindstone Pro includes an Impact Dashboard that uses a separate LLM to calculate time and money savings. The system is conservatively calibrated to avoid inflated estimates. The dashboard is isolated from individual workspaces, allowing IT leaders to evaluate adoption without surveilling private agent activity. This addresses a practical challenge: proving AI's value without invading privacy.
Fair Source licensing and real-world adoption
Rebel is released under a Fair Source license: the code is viewable, modifiable, and deployable. Organizations with fewer than 100 concurrent users can use it for free; beyond that, a commercial license is required. Additionally, after 24 months, a given version automatically converts to MIT open source. This reduces vendor dependency risk. Mindstone has already deployed Rebel at Epignosis, a 250-employee company, recovering the equivalent of eight full-time roles over 12 weeks. Adoption spread organically, a phenomenon dubbed the potatoes effect.
Sponsored Protocol
Mindstone raised $5 million from investors including Pearson Ventures and Moonfire Ventures. The platform is available for macOS and Windows, with Linux support in development. For those seeking an alternative to centralized cloud systems, Rebel offers a new way to orchestrate autonomous AI agents without sacrificing data sovereignty. To explore more AI productivity tools, read about Google Finance Launches Dedicated Android App with AI Features and Google Drive Organization for Businesses. For more on the Markdown format, see the Wikipedia page.
