🛰️ DSM: Dynamic Segmented Memory (MCP Server)

DSM is a high-performance Dynamic Segmented Memory engine implemented as an MCP (Model Context Protocol) server.

Unlike traditional vector databases, DSM organizes information into a hierarchy of segments, supports hybrid search, and enables multi-hop reasoning directly over your knowledge base.

🔌 What is MCP?

The Model Context Protocol (MCP) is an open standard that enables AI agents (like Claude Desktop, Cursor, or Windsurf) to securely connect to external data and tools.

By running DSM as an MCP server, you give your AI agent a persistent "long-term memory" that it can query, write to, and reason with, using a standardized set of tools.

🛠️ Features

• Hybrid Search: Seamlessly combines Dense (semantic) and Sparse (BM25) search for maximum retrieval precision.
• Reasoning: The dsm_reason tool allows agents to traverse the memory graph, finding deep connections instead of just keywords.
• Auto-Sync: One-click indexing of your entire project via dsm_sync.
• Conflict Control: Automatically detects contradictory information within the memory.

🚀 Quick Start

1. Install

pip install mcp dsm-memory

2. Configure MCP

Add the server to your claude_desktop_config.json or .claude.json:

{
  "mcpServers": {
    "dsm": {
      "command": "python",
      "args": ["path/to/mcp_server.py"],
      "env": {
          "DSM_PROJECT_ROOT": "/path/to/your/project"
      }
    }
  }
}

🔌 Tools

• dsm_search: Fast hybrid search over all memory segments.
• dsm_write: Manually commit new insights or architectural data.
• dsm_reason: Multi-hop recursive search for deep context.
• dsm_sync: Full codebase indexing and synchronization.
• dsm_info: Live statistics about the memory state.

📄 License

MIT ©