tverney/mcp-agent-memory

Overview

An MCP server that provides persistent memory for AI agents across MCP-compatible clients like Kiro, Claude Desktop, and Cursor. It stores and retrieves agent state using a daemon process, enabling long-running conversations and context retention.

Best for

Best for
Developers using MCP clients who need persistent agent memory across sessions

Use cases

• Maintain agent memory across Claude Desktop sessions
• Persist conversation context in Cursor for ongoing coding tasks
• Share agent state between different MCP clients

How to use

Install

npm install -g mcp-agent-memory

Tools exposed

• MEMORY_DIRECTORY
• SESSION_DIRECTORY

Tested with

Claude Desktop, Cursor, ChatGPT

Example client config

{}

Notes

An MCP server that provides persistent memory for AI agents across MCP-compatible clients like Kiro, Claude Desktop, and Cursor. It stores and retrieves agent state using a daemon process, enabling long-running conversations and context retention.

3 stars on GitHub. Last updated 2026-05-23. Licensed MIT.

Use cases

• Maintain agent memory across Claude Desktop sessions
• Persist conversation context in Cursor for ongoing coding tasks
• Share agent state between different MCP clients

Pros

• Works with multiple popular MCP clients out of the box
• Simple setup as a standalone MCP server
• Enables persistent memory without client-specific hacks

Cons

• Requires running a separate daemon process
• Limited to MCP-compatible clients only
• Small community with only 3 GitHub stars

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.