agent-memory-mcp

To store persistent memory for AI agents, you use an MCP server with memory_write and memory_read capabilities. This allows your agents to save and retrieve architecture decisions, patterns, and documentation across long-running projects.

An MCP server memory bank provides persistent, searchable memory for AI agents. It uses memory_search, memory_write, memory_read, and memory_stats functions to maintain and query long-term knowledge and project documentation across sessions.

Yes, you need Node.js v18+ to set up persistent searchable memory for your AI agent. After installing Node.js, you must clone the repository, run npm install, compile the code, and start the MCP server.

Yes, you can visualize and search stored AI agent memories using a standalone dashboard. The memory dashboard runs at http://localhost:3333 and provides analytics for your stored memories and project decisions.

The best way to prevent AI agents from forgetting architecture decisions is implementing a persistent memory bank. This automatically syncs with project documentation and uses an MCP server to store and retrieve searchable knowledge.

You can check the memory stats and knowledge capacity of your AI agent memory server using the built-in memory_stats function. This provides analytics on stored memories and decisions within your persistent memory bank.