Mirix is a multi-agent personal assistant that builds and manages memory through continuous screen activity tracking and natural conversation.
The tool addresses the challenge of retaining and organizing information from digital experiences by capturing real-time visual data and consolidating it into structured memories. It employs six specialized memory components—Core, Episodic, Semantic, Procedural, Resource, and Knowledge Vault—each managed by dedicated agents. This architecture allows the system to transform raw screen observations into a queryable knowledge base that adapts to your workflow. The system supports multi-modal input including text, images, voice, and screen captures, and provides an auto-dream consolidation feature that reviews existing memories, merges duplicates, and resolves conflicts to keep the knowledge base clean and coherent.
Mirix suits developers and users who want a privacy-first personal AI assistant that learns from their digital activities without sending long-term data to external servers. All persistent data is stored locally with user-controlled privacy settings. The tool is accessible through a Python client library and exposes REST endpoints for integration. It uses PostgreSQL-native BM25 full-text search combined with vector similarity for advanced memory retrieval. Those evaluating adoption should be aware that the project requires Docker for the backend and dashboard setup, and that existing databases need manual migration scripts when upgrading to versions that alter the schema, though the server will identify which migrations are required.
The project maintains active development with regular updates to its core memory system and procedural memory capabilities. Documentation is comprehensive, including an academic paper describing the approach. The team provides multiple channels for community engagement including Discord and maintains clear communication about database upgrade requirements and migration procedures.
