Virtual Filesystem

关于

This tool addresses the common challenge of data loss in AI agent runtimes by decoupling storage from the agent's execution environment. It offers a PostgreSQL-backed virtual filesystem that ensures file operations are persistent, surviving process restarts and scaling across deployments. Each agent session automatically receives its own isolated namespace, while named persistent stores enable long-term agent memory across sessions. By exposing standard POSIX-like file operations through Model Context Protocol (MCP) tools, it empowers AI agents with robust and reliable context management.

主要功能

• Automatic session isolation for each agent
• 11 standard POSIX-style file operation tools (read, write, ls, glob, grep, etc.)
• Persistent storage of files in PostgreSQL
• 0 GitHub stars
• Cross-session named stores for long-term agent memory
• Optional Row Level Security for database-enforced session isolation

使用案例

• Decoupling AI agent storage from runtime environments to prevent data loss on container restarts
• Managing context and scratchpad data for AI agents
• Providing persistent, long-term memory for AI agents across sessions