odysseus-dev/odysseus

Odysseus is a self-hosted AI workspace that integrates chat, agents, research, documents, email, notes, and calendar functionality with support for local and API-based models.

The tool addresses the need for a unified, privacy-preserving environment where users can run AI workflows without relying on external services. It combines multiple capabilities—chat with local or API models, multi-step web research with source reading, side-by-side model comparison, a writing-first document editor with AI assistance, email management with IMAP/SMTP, task scheduling, and calendar synchronization—into a single self-hosted application. The workspace supports Model Context Protocol (MCP), shell execution, file handling, and skill-based extensibility, allowing users to build custom agent workflows.

Odysseus suits developers and organizations prioritizing data privacy and control over their AI infrastructure. It works well for teams wanting to consolidate multiple AI tools into one interface, those experimenting with local model serving, and users who need email integration alongside AI capabilities. The project emphasizes hardware-aware model recommendations and serves as both a chat interface and a research platform. The setup guide covers native installs, GPU configuration, and deployment across Windows, macOS, and Linux, with Docker Compose as the default deployment method. The README distinguishes the tool by highlighting its writing-first document editor, deep research capabilities with multi-step workflows, and integrated email triage rather than comparing it directly to other platforms.

The project maintains two branches with different stability profiles, with the dev branch receiving changes first and main offering a more curated experience. Development activity shows responsiveness to fresh-install testing and provider setup issues, with the roadmap and contributing guide indicating ongoing refinement of mobile interfaces, editor polish, and documentation. The maintainers actively solicit help on focused areas including small refactors and setup-related bugs, suggesting a project in active iteration rather than maintenance mode.