Prism
Enhances AI agents with production-ready persistent session memory, multi-engine search, and progressive context loading through a Model Context Protocol server.
Prism is a Model Context Protocol server designed to provide AI agents with advanced memory and knowledge management capabilities. It offers persistent session memory, progressive context loading, and multi-engine search, enabling AI agents in clients like Claude Desktop and Cursor to retain context across sessions and access a comprehensive knowledge base. This eliminates the need for repeated explanations, facilitates knowledge accumulation, supports sophisticated information retrieval, and allows for efficient analysis for complex tasks.
주요 기능
013 GitHub stars
02MCP Prompts and Resources for zero-tool-call context injection before LLM thinking
03Optimistic Concurrency Control (OCC) with version tracking and multi-tenant Row Level Security (RLS)
04Gemini-powered ledger compaction and semantic search via pgvector embeddings
05Persistent Session Memory with progressive context loading (quick, standard, deep levels)
06Multi-engine search combining Brave Search, Vertex AI Discovery Engine, and Gemini analysis
사용 사례
01Facilitating efficient knowledge transfer and project onboarding for new team members or clients.
02Managing context for long-running AI agent development projects across multiple sessions.
03Enabling multi-agent collaboration with shared, project-isolated memory via Supabase.