home / mcp / memory mcp server
Provides local, durable, searchable memory for Claude Desktop and MCP-compatible assistants with optional cloud sync.
Configuration
View docs{
"mcpServers": {
"whenmoon-afk-claude-memory-mcp": {
"command": "npx",
"args": [
"-y",
"github:whenmoon-afk/claude-memory-mcp"
],
"env": {
"MEMORY_DB_PATH": "Platform-specific database file location",
"DEFAULT_TTL_DAYS": "90"
}
}
}
}Memory MCP provides local, persistent memory storage for Claude Desktop and MCP-compatible AI assistants. It runs on your machine, delivering durable, searchable memory with fast full-text search, entity extraction, and summarization, while keeping data under your control.
You connect to the Memory MCP server from an MCP client by configuring a memory MCP entry as a locally running service. Use the provided npx-based command to start the memory MCP, and then point your MCP client to that local server. Once connected, you can store memories with automatic summarization, search memories with token-aware loading, and soft-delete memories while preserving provenance.
Prerequisites: you need Node.js 18 or newer installed on your system. You also need npm or npx available in your environment.
1. Install the memory MCP by running the CLI command shown in the manual configuration option below.
2. Restart Claude Desktop after installation to ensure the new MCP is loaded.
Configuration and usage details are provided through the Merger Cloud configuration option and the local MCP server command. This section covers how to enable local memory storage, how to control memory expiration, and how to manage data privacy.
Security and privacy: Memory data is stored locally by default. Cloud sync is optional and requires explicit configuration. Local mode collects zero telemetry by default, while cloud mode captures memory content you choose to sync and usage metrics for service improvement.
Troubleshooting and notes: If you encounter issues, ensure Claude Desktop is fully restarted after changes, verify the JSON config syntax if you opt for manual configuration, and confirm Node.js version 18+ is installed. If connection issues persist on Windows, use the dedicated startup wrapper configuration described below.
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["-y", "github:whenmoon-afk/claude-memory-mcp"]
}
}
}Windows users may need to use a full command wrapper, for example: "cmd" with arguments ["/c", "npx", "-y", "github:whenmoon-afk/claude-memory-mcp"] to ensure proper startup on that platform.
- memory_store: Store a memory with auto-summarization and entity extraction - memory_recall: Search memories with token-aware loading - memory_forget: Soft-delete a memory (preserves audit trail)
- MEMORY_DB_PATH: Database file location (default platform-specific path) - DEFAULT_TTL_DAYS: Default memory expiration (default 90 days)
Tools not appearing in Claude Desktop? Restart Claude Desktop completely, verify the config file is valid JSON, and ensure Node.js 18+ is installed.
Connection issues on Windows may require using the cmd wrapper or the full path to npx.cmd.
If you experience stale cached versions, consider clearing npm cache or using a global install approach for version control. Startup with the github: method may take longer on first run due to dependency installation.
- @modelcontextprotocol/sdk — MCP protocol implementation - better-sqlite3 — Fast native SQLite with FTS5
This memory MCP server runs locally and is designed for Claude Desktop and MCP-compatible assistants.
Local-only by default: Memories are stored locally at the default path and are not sent anywhere unless you enable cloud sync. Cloud sync is optional and governed by your Substratia account and privacy settings. The project is MIT-licensed.
Store a memory with auto-summarization and entity extraction
Search memories with token-aware loading
Soft-delete a memory while preserving audit trails