Zettelkasten MCP Server

Example Use Cases

• 1

  Persist and retrieve cross-session conversational context using semantic embeddings and graph relationships.

• 2

  A local stdio MCP transport to integrate with desktop assistants like Claude or a networked SSE transport for browser-based clients such as Cursor.

• 3

  Scalable, hybrid storage (Pinecone + Neo4j) to combine fast similarity search with relationship-aware queries for richer context retrieval.

Description

Zettelkasten is an MCP server designed for chatbot applications that combines a vector database (Pinecone) and a graph database (Neo4j) to store and retrieve contextual chat memory across sessions. It supports two transports—stdio and Server-Sent Events (SSE)—so it can integrate with local clients (e.g., Claude desktop) or networked apps (e.g., Cursor). The hybrid architecture enables semantic similarity search via embeddings and relationship-based queries via a graph, improving context-aware and scalable conversational agents. The project is distributed as source code and requires external services and API keys (Pinecone, Neo4j, OpenAI) to operate.