Persistent MCP server for Claude Code, self-hosted

Persistent MCP Server for Claude Code

This repository provides a self-hosted MCP server specifically designed for Claude Code. It leverages a robust stack including n8n for workflow automation, PostgreSQL for relational data storage, and pgvector for efficient similarity search and persistent memory. This setup allows developers to maintain state and context for their AI agents without relying on cloud-based solutions, offering greater control and privacy.

What it Does

The Persistent MCP Server acts as a backend for Claude Code agents, enabling them to retain information across interactions. By integrating with PostgreSQL and pgvector, it stores conversation history, learned patterns, and other relevant data. This persistent memory allows Claude Code agents to build upon previous interactions, leading to more coherent and context-aware responses. n8n orchestrates the data flow, managing triggers, actions, and integrations within the server environment.

Key Features

• Self-Hosted: Full control over your data and infrastructure.
• Persistent Memory: Utilizes PostgreSQL and pgvector for long-term data storage and retrieval.
• n8n Integration: Powerful workflow automation for managing agent logic and data pipelines.
• Claude Code Optimized: Tailored for use with Claude Code agents.
• Developer Focused: Designed for technical users seeking robust AI agent infrastructure.

Who it's For

This tool is ideal for AI developers and engineers who require a private, self-hosted solution for building and deploying Claude Code agents. It's particularly suited for projects where data privacy is paramount, or where the cost and complexity of cloud-based AI memory solutions are a concern. Developers looking to experiment with advanced agent capabilities, such as long-term memory and contextual understanding, will find this persistent MCP server invaluable.