Knowledge Graph Memory

This Memory MCP Server provides a persistent local knowledge graph that helps your assistant remember user information across chats. It supports entities, relations, and observations, with powerful search, traversal, and filtering to keep memory accurate and efficient for personalized interactions.

How to use

You interact with the Memory MCP Server through an MCP client that runs as a separate process connected via the MCP protocol. You can create entities and observations, build relationships, read and search the graph, and traverse connections to understand how things relate. Use memory to store user identifiers, preferences, and contextual facts so the assistant can personalize conversations, recall important details, and surface relevant history when needed.

How to install

Prerequisites: Install Node.js on your system. You will also need access to the project files that expose the Memory MCP Server as a local process.

Prerequisites
- Node.js installed on your machine
- Access to the Memory MCP Server files (local project folder)

Install dependencies and build the server, then run the local process using the provided executable path.

npm install
npm run build
node C:/path/to/better-memory-mcp/dist/index.js

Configuration and usage notes

Configure your client to connect to the Memory MCP Server as a local stdio server using the command and environment shown in the usage example. The server reads its memory storage location from an environment variable.

Example environment setup (Windows path shown as in an example): set MEMORY_FILE_PATH=C:/path/to/your/memory.jsonl

Additional considerations

Security: Keep the memory data file in a secure location and restrict access to the host running the MCP server. Regularly review the observations to prune outdated or sensitive information. Performance: Use targeted searches and observation filtering to minimize memory scanning costs as your graph grows.

Available tools

create_entities

Create multiple new entities with a name, type, and initial observations.

create_relations

Create multiple directed relations between existing entities.

add_observations

Attach new observations to existing entities and return what was added.

delete_entities

Remove entities and their related connections; cascading deletion of relations.

delete_observations

Remove specific observations from entities.

delete_relations

Remove specific relations between entities.

read_graph

Fetch the full knowledge graph with all entities and relations.

search_nodes

Advanced search for entities with boolean operators, fuzzy matching, and relevance scoring.

open_nodes

Retrieve specific nodes by name along with their inter-node relations.

search_observations

Search observations to return individual matching facts rather than full entities.

get_neighbors

Return all directly connected entities for a given entity, with direction and type filtering.

find_path

Find the shortest path between two entities using BFS, including path length.

get_subgraph

Extract a neighborhood around seed entities up to a specified depth.

filter_by_type

Return all entities of a specific type with their internal relations.

filter_relations

Filter relations by type, source, or target and return matching connections.

filter_observations

Find observations matching a pattern or preset categories like dated, techdebt, deprecated.