home / mcp / hippocampus memory mcp server

Hippocampus Memory MCP Server

Python-based MCP server that stores, retrieves, consolidates, and forgets memories via semantic search for persistent LLM context.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "jameslovespancakes-memory-mcp": {
      "command": "python",
      "args": [
        "-m",
        "memory_mcp_server.server"
      ]
    }
  }
}

You can run a Python-based MCP server that gives large language models persistent, hippocampus-inspired memory across sessions. It stores, retrieves, consolidates, and forgets memories using semantic similarity search, helping you build richer, long-lived context for your LLM workflows.

How to use

Connect to the memory MCP server from your MCP client to store and retrieve memories by semantic meaning. Use memory_write to persist new memories with tags and metadata, memory_read to fetch memories that match a query semantically, memory_consolidate to merge similar memories and reduce redundancy, memory_forget to remove memories by age, importance, or tags, and memory_stats to check system status.

How to install

Prerequisites: Python 3.9 or newer must be installed on your system.

pip install hippocampus-memory-mcp

Additional setup and configuration

Start a local MCP server using the provided development command. This runs the server in your environment without needing a remote host.

python -m memory_mcp_server.server

Note on integration with Claude Desktop

To add persistent memory support to Claude Desktop, configure the MCP server entry so Claude launches the server as a local process. The example below shows how to invoke the server directly from Python.

{
  "mcpServers": {
    "memory": {
      "command": "python",
      "args": ["-m", "memory_mcp_server.server"]
    }
  }
}

Available tools

memory_read

Retrieve memories by semantic similarity using a query text and optional filters.

memory_write

Store new memories with text, tags, metadata, and an optional importance score.

memory_consolidate

Merge similar memories to reduce redundancy based on a similarity threshold.

memory_forget

Remove memories by age, importance, or tags to manage storage.

memory_stats

Provide system statistics and memory usage metrics.