Memory Custom
Extends the MCP Memory server to create and manage a knowledge graph from LLM interactions.
Memory Custom
This project adds new features to the Memory server offered by the MCP team. It allows for the creation and management of a knowledge graph that captures interactions via a language model (LLM).
New Features
1. Custom Memory Paths
- Users can now specify different memory file paths for various projects.
- Why?: This feature enhances organization and management of memory data, allowing for project-specific memory storage.
2. Timestamping
- The server now generates timestamps for interactions.
- Why?: Timestamps enable tracking of when each memory was created or modified, providing better context and history for the stored data.
Getting Started
Prerequisites
- Node.js (version 16 or higher)
Installing via Smithery
To install Knowledge Graph Memory Server for Claude Desktop automatically via Smithery:
npx -y @smithery/cli install @BRO3886/mcp-memory-custom --client claude
Installation
-
Clone the repository:
git clone git@github.com:BRO3886/mcp-memory-custom.git cd mcp-memory-custom -
Install the dependencies:
npm install
Configuration
Before running the server, you can set the MEMORY_FILE_PATH environment variable to specify the path for the memory file. If not set, the server will default to using memory.json in the same directory as the script.
Running the Server
Updating the mcp server json file
Add this to your claude_desktop_config.json / .cursor/mcp.json file:
{
"mcpServers": {
"memory": {
"command": "node",
"args": ["/path/to/mcp-memory-custom/dist/index.js"]
}
}
}
System Prompt changes:
Follow these steps for each interaction:
1. The memoryFilePath for this project is /path/to/memory/project_name.json - always pass this path to the memory file operations (when creating entities, relations, or retrieving memory etc.)
2. User Identification:
- You should assume that you are interacting with default_user
- If you have not identified default_user, proactively try to do so.
3. Memory Retrieval:
- Always begin your chat by saying only "Remembering..." and retrieve all relevant information from your knowledge graph
- Always refer to your knowledge graph as your "memory"
4. Memory
- While conversing with the user, be attentive to any new information that falls into these categories:
a) Basic Identity (age, gender, location, job title, education level, etc.)
b) Behaviors (interests, habits, etc.)
c) Preferences (communication style, preferred language, etc.)
d) Goals (goals, targets, aspirations, etc.)
e) Relationships (personal and professional relationships up to 3 degrees of separation)
5. Memory Update:
- If any new information was gathered during the interaction, update your memory as follows:
a) Create entities for recurring organizations, people, and significant events, add timestamps to wherever required. You can get current timestamp via get_current_time
b) Connect them to the current entities using relations
c) Store facts about them as observations, add timestamps to observations via get_current_time
IMPORTANT: Provide a helpful and engaging response, asking relevant questions to encourage user engagement. Update the memory during the interaction, if required, based on the new information gathered (point 4).
Running the Server Locally
To start the Knowledge Graph Memory Server, run:
npm run build
node dist/index.js
The server will listen for requests via standard input/output.
API Endpoints
The server exposes several tools that can be called with specific parameters:
- Get Current Time
- Set Memory File Path
- Create Entities
- Create Relations
- Add Observations
- Delete Entities
- Delete Observations
- Delete Relations
- Read Graph
- Search Nodes
- Open Nodes
Acknowledgments
- Inspired by the Memory server from Anthropic.
Related Servers
Redshift MCP Server
An MCP server for Amazon Redshift, allowing AI assistants to interact with Redshift databases.
Graphiti MCP Server
A framework for building and querying temporally-aware knowledge graphs for AI agents.
Blackbaud FE NXT by CData
A read-only MCP server for Blackbaud FE NXT by CData, enabling LLMs to query live data. Requires a separate CData JDBC Driver.
Coresignal
Access comprehensive B2B data on companies, employees, and job postings for your LLMs and AI workflows.
SupaMCP Server
A runtime-configurable MCP server that turns a Supabase project into an AI-compatible tool interface.
Hive MCP Server
Enables AI assistants to interact with the Hive blockchain through the Model Context Protocol.
Fedspeak MCP Server
Access and analyze Federal Reserve (FOMC) statements.
Data Mesh Manager MCP
Discover data products and request access in Data Mesh Manager.
RentCast
Access property data, valuations, and market statistics using the RentCast API.
CData Bing Ads
A read-only MCP server to query live Bing Ads data using CData's JDBC driver.