Developer MicrotronX has released mxLore, a self-hosted MCP server designed to give AI coding assistants persistent memory across projects, sessions, and teams. Instead of starting from scratch every time, AI tools can recall architectural decisions, specifications, plans, lessons learned, and project knowledge accumulated over time.
mxLore acts as a centralized knowledge layer for Claude Code, Cursor, Windsurf, claude.ai, and other MCP-compatible clients, allowing AI assistants to work with the same shared project context regardless of where they’re being used.
The framework includes support for:
- Persistent AI memory
- MCP-based AI assistant integration
- Multi-project knowledge management
- Semantic and full-text search
- Institutional memory and lessons learned
- Multi-agent communication
- AI skills and orchestration workflows
- Team collaboration and administration tools
Built entirely in Delphi, mxLore combines TMS Sparkle, FireDAC, and MariaDB to deliver a self-hosted knowledge platform with dozens of MCP tools for storing, searching, and retrieving development knowledge. The system is designed to help AI assistants understand not just your code, but the reasoning, decisions, and context behind it.
Whether you’re using Claude Code, Cursor, or your own AI agents, mxLore provides the long-term memory layer that modern AI-assisted development workflows have been missing.