Simple Memory MCP Server

Simple Memory MCP Server

A lightweight MCP server that provides persistent knowledge graph storage for AI assistants, enabling memory across sessions through entity-relationship storage with JSON file persistence.

Category
Visit Server

README

Simple Memory MCP Server

A lightweight Model Context Protocol (MCP) server that provides persistent knowledge graph storage for AI assistants. Enables AI agents to maintain memory across sessions through entity-relationship storage with JSON file persistence.

๐Ÿš€ Features

  • Persistent Memory: Knowledge graph storage with automatic persistence to JSON files
  • Entity Management: Create, read, update, and delete entities with typed observations
  • Relationship Tracking: Manage relationships between entities with type annotations
  • Search Capabilities: Full-text search across entity names, types, and observations
  • MCP Compliant: Full Model Context Protocol v2025-06-18 compatibility
  • Simple Architecture: Lightweight, single-file implementation with minimal dependencies

๐Ÿ“‹ Table of Contents

๐Ÿ›  Installation

Interactive Installation (Recommended)

# Bash installer with interactive configuration
curl -fsSL https://raw.githubusercontent.com/your-username/simple-memory-mcp/main/install.sh | bash

What it does:

  • ๐Ÿ” Auto-detects your Obsidian vaults
  • ๐Ÿ“ Configures custom memory storage location
  • โš™๏ธ Sets up Claude Desktop/Cursor automatically
  • ๐Ÿ—‚๏ธ Optional Obsidian auto-export configuration

Interactive Setup Flow:

๐Ÿ“ Memory Storage Configuration
Where should memory be stored? [~/.cursor/memory.json]: 

๐Ÿ—‚๏ธ Obsidian Integration  
Do you use Obsidian? (y/n) [n]: y

๐Ÿ“š Found Obsidian vaults:
  1. My Knowledge Base (/Users/you/Documents/MyVault)
  2. Work Notes (/Users/you/Desktop/WorkVault)
Choose vault (1-2) or enter custom path [1]: 1

Enable auto-export after entity creation? (y/n) [n]: y
Export format (markdown/dataview/canvas/all) [markdown]: all

Manual Installation

Prerequisites

  • Node.js v18.x or higher
  • npm or pnpm package manager

Install Dependencies

npm install

Environment Setup

The server automatically saves memory to:

  • ~/.cursor/memory.json (default)
  • Custom path via MEMORY_PATH environment variable
# Optional: Set custom memory file location
export MEMORY_PATH="/path/to/your/memory/directory"

๐Ÿš€ Quick Start

1. Start the Server

npm start
# or
node index.js

2. Test with MCP Inspector

# Install and run MCP Inspector
npx @modelcontextprotocol/inspector

# Configure server in Inspector:
# Command: node
# Args: /path/to/your/simple-memory-mcp/index.js

3. Basic Usage Example

// Create entities
await client.callTool({
  name: "create_entities",
  arguments: {
    entities: [{
      name: "john-doe",
      entityType: "person",
      observations: ["Software engineer", "Works remotely", "Enjoys hiking"]
    }]
  }
});

// Create relationships
await client.callTool({
  name: "create_relations",
  arguments: {
    relations: [{
      from: "john-doe",
      to: "acme-corp",
      relationType: "works_for"
    }]
  }
});

// Search entities
await client.callTool({
  name: "search_nodes",
  arguments: {
    query: "engineer"
  }
});

๐Ÿ“š API Reference

Tools Overview

Tool Description Input Output
create_entities Create multiple entities {entities: Entity[]} Created entities
create_relations Create relationships {relations: Relation[]} Created relations
add_observations Add observations to entities {observations: Observation[]} Updated observations
delete_entities Delete entities and relations {entityNames: string[]} Deleted entities
delete_observations Remove specific observations {deletions: Deletion[]} Deleted observations
delete_relations Remove relationships {relations: Relation[]} Deleted relations
read_graph Get complete knowledge graph {} Full graph data
search_nodes Search entities by query {query: string} Matching entities
open_nodes Get specific entities {names: string[]} Requested entities
export_to_obsidian Export graph to Obsidian vault {vaultPath: string, format?: string} Export result

Data Types

Entity

interface Entity {
  name: string;           // Unique identifier
  entityType: string;     // Type classification
  observations: string[]; // Array of observation texts
}

Relation

interface Relation {
  from: string;          // Source entity name
  to: string;            // Target entity name
  relationType: string;  // Relationship type
}

Observation

interface Observation {
  entityName: string;    // Target entity name
  contents: string[];    // New observations to add
}

Deletion

interface Deletion {
  entityName: string;      // Target entity name
  observations: string[];  // Observations to remove
}

Detailed Tool Documentation

create_entities

Creates multiple new entities in the knowledge graph.

Input Schema:

{
  "entities": [
    {
      "name": "entity-name",
      "entityType": "person|organization|concept|etc",
      "observations": ["observation1", "observation2"]
    }
  ]
}

Example:

{
  "entities": [
    {
      "name": "alice-johnson",
      "entityType": "person",
      "observations": ["Data scientist", "PhD in Computer Science", "Lives in San Francisco"]
    },
    {
      "name": "tech-startup-xyz",
      "entityType": "organization", 
      "observations": ["AI/ML company", "Founded in 2023", "Series A funding"]
    }
  ]
}

Response:

[
  {
    "name": "alice-johnson",
    "entityType": "person",
    "observations": ["Data scientist", "PhD in Computer Science", "Lives in San Francisco"]
  }
]

create_relations

Creates relationships between existing entities.

Input Schema:

{
  "relations": [
    {
      "from": "source-entity",
      "to": "target-entity", 
      "relationType": "relationship-type"
    }
  ]
}

Example:

{
  "relations": [
    {
      "from": "alice-johnson",
      "to": "tech-startup-xyz",
      "relationType": "works_for"
    }
  ]
}

search_nodes

Search entities using full-text search across names, types, and observations.

Input Schema:

{
  "query": "search-term"
}

Example:

{
  "query": "data scientist"
}

Response: Array of matching entities with complete data.

read_graph

Returns the complete knowledge graph with all entities and relations.

Input Schema:

{}

Response:

{
  "entities": [
    {
      "name": "alice-johnson",
      "entityType": "person",
      "observations": ["Data scientist", "PhD in Computer Science"]
    }
  ],
  "relations": [
    {
      "from": "alice-johnson",
      "to": "tech-startup-xyz", 
      "relationType": "works_for"
    }
  ]
}

export_to_obsidian

Export the knowledge graph to an Obsidian vault in various formats.

Input Schema:

{
  "vaultPath": "/path/to/obsidian/vault",
  "format": "markdown",
  "autoIndex": true
}

Parameters:

  • vaultPath (required): Path to the Obsidian vault directory
  • format (optional): Export format - "markdown", "dataview", "canvas", or "all" (default: "markdown")
  • autoIndex (optional): Whether to create index files (default: true)

Example:

{
  "vaultPath": "/Users/username/Documents/MyVault",
  "format": "all",
  "autoIndex": true
}

Response:

{
  "success": true,
  "vaultPath": "/Users/username/Documents/MyVault",
  "format": "all",
  "entityCount": 42,
  "relationCount": 18,
  "timestamp": "2024-01-15T10:30:00.000Z"
}

โš™๏ธ Configuration

Environment Variables

Variable Default Description
MEMORY_PATH ~/.cursor/memory.json Custom memory file location
NODE_ENV development Runtime environment
OBSIDIAN_AUTO_EXPORT false Enable automatic Obsidian export after entity creation
OBSIDIAN_VAULT_PATH - Path to Obsidian vault for auto-export
OBSIDIAN_EXPORT_FORMAT markdown Export format for auto-export

Memory File Structure

The server persists data in JSON format:

{
  "entities": [
    {
      "name": "entity-name",
      "entityType": "type",
      "observations": ["obs1", "obs2"]
    }
  ],
  "relations": [
    {
      "from": "entity1",
      "to": "entity2",
      "relationType": "relationship"
    }
  ]
}

MCP Client Configuration

For Claude Desktop, add to your MCP settings:

{
  "mcpServers": {
    "simple-memory": {
      "command": "node",
      "args": ["/path/to/simple-memory-mcp/index.js"],
      "env": {
        "MEMORY_PATH": "/custom/path/to/memory/directory"
      }
    }
  }
}

๐Ÿงช Testing

Running Tests

# Run comprehensive server test
node test-server.js

Expected Test Output

๐Ÿงช Testing Simple Memory MCP Server...
โœ… Connected successfully!
โœ… Found 9 tools: create_entities, create_relations, ...
โœ… All tests passed! Server is working correctly.

Manual Testing with Inspector

  1. Start MCP Inspector: npx @modelcontextprotocol/inspector
  2. Configure server connection
  3. Test each tool with sample data
  4. Verify persistence by restarting server

Integration Testing

Test with actual MCP clients:

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "node",
  args: ["index.js"]
});

const client = new Client({
  name: "test-client",
  version: "1.0.0"
}, {
  capabilities: {}
});

await client.connect(transport);

๐Ÿ”ง Troubleshooting

Common Issues

Server Won't Start

Error: Cannot read properties of undefined (reading 'method')

Solution: Ensure you're using the correct MCP SDK version and schema imports:

import {
  ListToolsRequestSchema,
  CallToolRequestSchema,
  ListPromptsRequestSchema,
  ListResourcesRequestSchema
} from '@modelcontextprotocol/sdk/types.js';

Missing Capabilities Error

Error: Server does not support prompts (required for prompts/list)

Solution: Declare all capabilities in server configuration:

const server = new Server(
  { name: 'simple-memory-mcp', version: '1.1.0' },
  {
    capabilities: {
      tools: {},
      prompts: {},
      resources: {}
    }
  }
);

Memory File Permissions

Error: EACCES: permission denied

Solution: Ensure write permissions to memory directory:

mkdir -p ~/.cursor
chmod 755 ~/.cursor

Tool Not Found

Error: Unknown tool: create_entities

Solution: Verify tool registration matches the schema names exactly.

Debug Mode

Enable detailed logging:

console.error("Debug info:", JSON.stringify(data, null, 2));

Performance Issues

For large knowledge graphs (>10,000 entities):

  1. Consider implementing pagination for read_graph
  2. Add indexing for search operations
  3. Implement lazy loading for entity details

๐Ÿ›  Development

Project Structure

simple-memory-mcp/
โ”œโ”€โ”€ index.js              # Main server implementation
โ”œโ”€โ”€ package.json          # Dependencies and scripts
โ”œโ”€โ”€ test-server.js        # Comprehensive test suite
โ”œโ”€โ”€ inspector-config.json # MCP Inspector configuration
โ”œโ”€โ”€ CLAUDE.md             # AI development protocols
โ””โ”€โ”€ README.md             # This documentation

Architecture

graph TD
    A[MCP Client] --> B[StdioServerTransport]
    B --> C[Simple Memory Server]
    C --> D[Entity Manager]
    C --> E[Relation Manager]
    C --> F[Search Engine]
    D --> G[JSON File Storage]
    E --> G
    F --> G

Core Classes

SimpleMemoryServer

Main server class handling:

  • Memory persistence (loadMemory(), saveMemory())
  • Entity operations (CRUD)
  • Relationship management
  • Search functionality

Key Methods:

  • createEntities(entities) - Batch entity creation
  • createRelations(relations) - Relationship creation
  • searchNodes(query) - Full-text search
  • readGraph() - Complete graph export

Extending the Server

Adding New Tools

  1. Define tool schema in tools/list handler
  2. Implement logic in tools/call handler
  3. Add method to SimpleMemoryServer class
  4. Update documentation

Custom Storage Backends

Replace JSON file storage:

class DatabaseMemoryServer extends SimpleMemoryServer {
  async saveMemory() {
    // Custom database implementation
  }
  
  async loadMemory() {
    // Custom database loading
  }
}

Contributing

  1. Fork the repository
  2. Create feature branch: git checkout -b feature-name
  3. Run tests: node test-server.js
  4. Commit changes: git commit -m "Description"
  5. Push branch: git push origin feature-name
  6. Create Pull Request

๐Ÿ“„ License

MIT License - see LICENSE file for details.

๐Ÿ“š Additional Documentation

๐Ÿค Support


Built with โค๏ธ using the Model Context Protocol

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured