AI-first knowledge base and "startup OS" designed to overcome the limitations of ephemeral AI sessions and scattered project data. It targets developers and power users seeking a self-hosted, data-portable solution where AI agents manage tasks and knowledge compounds over time, offering an "AI team" that remembers and grows without vendor lock-in.

How It Works

Cabinet operates on a "files on disk" philosophy, storing all data as Markdown files without a database, ensuring simplicity and portability. Its core innovation is an "AI-first" approach where AI agents with memory, goals, and scheduled jobs collaborate within the knowledge base. This enables compounding knowledge and automated task execution, mimicking a dynamic AI team. Key principles guiding its design include user data ownership ("Yours"), version control ("Git everything"), flexibility in AI providers ("BYOAI"), simplicity ("KISS"), security, and self-hosting.

Quick Start & Requirements

• Primary install/run command: npx create-cabinet@latest, cd cabinet, npm run dev:all.
• Prerequisites: Node.js 20+, Claude Code CLI (npm install -g @anthropic-ai/claude-code), macOS or Linux (Windows via WSL).
• Setup: "From zero to AI team in 2 minutes" via an onboarding wizard.
• Links: runcabinet.com, Discord community.

Highlighted Details

• AI agent orchestration with defined goals, skills, and scheduled cron jobs for automated workflows.
• Embedded HTML applications directly within the knowledge base, version-controlled via Git, offering unique integration possibilities.
• Git-backed version history for all data, enabling inspection, restoration, and auditing of knowledge evolution.
• Integrated web terminal for in-browser command-line access.
• File-based storage (Markdown) ensures data portability and avoids vendor lock-in, differentiating it from tools like Notion.

Maintenance & Community

Cabinet is built and developed in public by Hila Shmuel, former Engineering Manager at Apple. The primary community hub is a Discord server, used for questions, feedback, and shaping product direction. Contributions are welcomed, with a recommendation to sync with the builder via Discord before coding to align with the rapidly evolving roadmap.

Licensing & Compatibility

• License type: MIT License.
• Compatibility: The MIT license is permissive for commercial use and integration into closed-source projects. Its self-hosted, file-based nature ensures broad data compatibility.

Limitations & Caveats

The system is primarily designed for macOS or Linux environments, with Windows support contingent on using WSL. The project is noted to be in a rapid development phase, and contributors are advised to align with the core developer before initiating significant work to ensure coherence. While supporting "Bring Your Own AI," initial setup requires specific command-line tools.