Quoroom

" The platform enables the creation of a "room," which is a self-governing collective of agents that can run locally on macOS, Windows, or Linux, or scale to cloud environments. Each room starts with a queen agent that decomposes high-level objectives into sub-goals and delegates tasks to specialized worker agents. The agents coordinate through decentralized decision-making, employing quorum voting to propose, deliberate, and select actions, with the ability to modify their own voting rules as needed.

Quoroom emphasizes transparency and control, offering features such as full execution logs, checkpoints, and audit trails for every agent decision. Users can allocate budgets to allow agents to rent cloud servers or purchase services, with every transaction tracked through a dedicated payment account for each room. The system supports both private and public rooms, allowing users to choose whether their goals, decisions, and financials are visible to others. Agents can self-improve by refining their skills, prompts, and strategies over time, with safety measures like audit logs and the ability to revert changes. Cross-room intelligence is also supported, enabling rooms to observe, learn from, and negotiate with each other.

The platform includes a built-in assistant called Clerk, which can act as a personal AI assistant accessible via Telegram, email, or in-app chat. Clerk is capable of finding information, sending emails, setting reminders, and managing agents through natural language conversation, with persistent memory and web access. Quoroom supports integration with several AI models: a free local model is included, and users can bring their own API keys for OpenAI, Claude, and Gemini models, with pay-per-use options for these APIs. All data is stored locally in a SQLite database, ensuring user privacy unless cloud features are explicitly enabled.

Quoroom is distributed as open-source software under the MIT license, and no funding is required to start using it. Installation is available via Homebrew or npm, and the platform provides a local web UI and command-line interface for management. The tool is intended for research use, and users are responsible for monitoring agent actions and any associated API costs.