MindsHub Cowork is an open-source unified workspace platform designed to delegate complex tasks to AI agents and collect finished, shareable results. The repository serves as a superproject that integrates the desktop and web application, agent backend, and data engine into a single deployable stack. Users can delegate entire workflows—research, analysis, reporting, and scheduled operations—without writing code, and the platform transforms agent output into publishable artifacts like documents, dashboards, and applications.
The platform connects to external data sources through a secure credential vault that supports systems including BigQuery, PostgreSQL, MySQL, MSSQL, Gmail, Google Drive, HubSpot, Notion, and Linear. Credentials remain scoped per connection, ensuring agents never access raw API keys. A model router allows seamless switching between frontier models such as Claude, GPT, and Gemini alongside open-source alternatives like DeepSeek, Qwen, and Kimi without requiring separate key configuration for each provider. The system runs interchangeable open-source agent harnesses—Anton as the default and Hermes as an alternative—selectable from a dropdown interface.
The repository is structured as a Makefile-based superproject that pins submodules for the frontend, core API, core agent, and data vault to specific commits. Developers can work on feature branches using a gitignored dev.env configuration file, with make commands handling checkout and execution across all submodules. The build system supports multiple deployment modes: desktop application development with hot reload via make dev or make watch, web application development via make dev-web, production builds, and platform-specific packaging for macOS and Windows. A make flush command provides a clean slate by removing local runtime installations and application state.
The platform supports deployment across diverse infrastructure environments including cloud, VPC, on-premises, air-gapped, and hybrid setups, allowing organizations to maintain control over infrastructure, models, permissions, and data. Features include cross-session memory, a reusable skill library, and scheduled task execution. The project is released under the MIT License, with bundled components governed by their respective licenses. Access is available through a hosted web application at console.mindshub.ai, native desktop applications for macOS and Windows, or by building from source. A free tier is available with paid Pro plans offering frontier model access and private artifacts.
