Memory Custom

Memory Custom is an MCP server that extends the standard Model Context Protocol memory server to build and manage persistent knowledge graphs from language model conversations. It connects directly to local JSON storage files, allowing developers, researchers, and power users to retain structured context across sessions in clients like Claude Desktop and Cursor. The server introduces key enhancements over default memory implementations, notably the ability to specify custom file paths for individual projects and the addition of automatic interaction timestamps to track when observations or entities are created and updated. By organizing data into entities, relations, and observations, Memory Custom enables language models to systematically capture user identity details, recurring concepts, project goals, and relational structures. Users interact with the knowledge graph through dedicated tools that allow the assistant to search nodes, read graph contents, create or delete entities, and record dated observations. This architecture ensures that conversations maintain continuity without mixing memory across separate codebases or client workflows, delivering an adaptable contextual memory layer for long-term agentic interactions.

Category: AI Memory & Context

Tags: context, graph database, knowledge-graph, memory

Visit Memory Custom

How to install and configure Memory Custom

1. Ensure Node.js (version 16 or higher) is installed. 2. Install via Smithery by running: npx -y @smithery/cli install @BRO3886/mcp-memory-custom --client claude 3. Alternatively, clone the repository and install dependencies manually: bash git clone git@github.com:BRO3886/mcp-memory-custom.git cd mcp-memory-custom npm install npm run build 4. Add the server to your claude_desktop_config.json or .cursor/mcp.json file: json { "mcpServers": { "memory": { "command": "node", "args": ["/path/to/mcp-memory-custom/dist/index.js"] } } } 5. Optionally set the MEMORY_FILE_PATH environment variable to define a custom path, or pass the memory file path directly via system prompt instructions.

What you can do with Memory Custom

• Segregating conversational context and entities into separate JSON storage files across distinct software development projects or client workspaces. - Chronologically logging user preferences, habits, and background facts with timestamps using the get_current_time tool for temporal context. - Mapping multi-degree relationships between people, organizations, and events mentioned during brainstorming sessions into a persistent graph structure. - Querying past observations and reading entire graph structures to recall previous decisions before starting a new conversation. - Updating or pruning outdated knowledge graph entities and observations dynamically during interactive problem-solving workflows.

Key facts

• Open Source
• https://github.com/BRO3886/mcp-memory-custom
• AI Memory & Context, Notes & Knowledge Management
• context, graph database, knowledge-graph, memory

Part of MCP Servers

How do I install Memory Custom?

You can install Memory Custom automatically using Smithery by running the command npx -y @smithery/cli install @BRO3886/mcp-memory-custom --client claude. Alternatively, clone the repository from GitHub, run npm install, compile the source using npm run build, and register the resulting dist/index.js script in your MCP client configuration.

What can Memory Custom do?

Memory Custom allows your language model to store and retrieve contextual information in a local knowledge graph. It provides tools to create, search, and delete entities, relations, and observations. Additionally, it supports custom file paths for project-specific memory storage and provides timestamps to track when data is recorded.

Which MCP clients work with Memory Custom?

Memory Custom works with any MCP-compliant client that supports local standard input and output transport. It is explicitly designed for use with Claude Desktop and Cursor. Users configure it by pointing their client configuration files, such as claude_desktop_config.json or .cursor/mcp.json, to the compiled Node.js runtime script.

Is Memory Custom open source?

Yes, Memory Custom is open source. You can access the complete source code, review its implementation, submit contributions, or raise issues directly on its official GitHub repository at https://github.com/BRO3886/mcp-memory-custom. Node.js version 16 or higher is required to run and build the project locally.

How does Memory Custom differ from the standard MCP memory server?

Memory Custom introduces two primary enhancements over the official Anthropic Memory server. First, it allows users to specify custom memory file paths, enabling isolated memory graphs for distinct projects. Second, it adds interaction timestamping through a dedicated tool, ensuring stored observations and entities retain temporal context.