Quick Start Guide
Get up and running with memcp in just a few minutes.
1. Create Your Account
Sign in with Google or GitHub to create your personal memory system.
2. Generate an API Key
Visit the API Keys page to generate a key for your AI tools.
3. Connect Your AI Tool
Follow the integration guides in the "AI Integration" section below.
4. Create Your First Memory
Use the create_memory tool or the web interface to store your first piece of information.
Create a memory: "Remember that I prefer dark mode in all my applications" #preferences #ui
5. Search Your Memories
Use search_memories to find information instantly.
Search for: "dark mode preferences"
General
ChatGPT & Claude Code Web
Connect memcp to ChatGPT or Claude Code Web using OAuth authentication.
Add a new MCP server with these settings:
MCP Server URL:
https://memcp.xyz/api/mcpAuthentication:
OAuthOAuth Client ID:
YOUR_API_KEYOAuth Client Secret:
YOUR_API_KEY(Use the same API key for both fields)
OAuth Flow:
- You'll be redirected to memcp.xyz to authorize
- Sign in if not already logged in
- Authorize the connection
- You'll be redirected back to the client
Replace YOUR_API_KEY with an API key from your API Keys page.
Code Editors
Claude Desktop
Add memcp to your Claude Desktop configuration to access your memories in conversations.
Configuration file location:
~/Library/Application Support/Claude/claude_desktop_config.jsonAdd to your config:
{
"mcpServers": {
"memcp": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-fetch"],
"env": {
"FETCH_URL": "https://memcp.xyz/api/mcp",
"FETCH_HEADERS": "{\"Authorization\": \"Bearer YOUR_API_KEY\"}"
}
}
}
}Replace YOUR_API_KEY with an API key from your API Keys page.
Windsurf IDE
Integrate memcp with Windsurf for AI-assisted coding with persistent memory.
Add to Windsurf settings:
{
"mcp": {
"servers": {
"memcp": {
"url": "https://memcp.xyz/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
}Replace YOUR_API_KEY with an API key from your API Keys page.
Cursor IDE
Enable memcp in Cursor for context-aware AI assistance.
Configuration:
{
"mcpServers": {
"memcp": {
"url": "https://memcp.xyz/api/mcp",
"apiKey": "YOUR_API_KEY"
}
}
}Replace YOUR_API_KEY with an API key from your API Keys page.
Any MCP-Compatible Client
memcp implements the Model Context Protocol and works with any compatible client.
Endpoint:
https://memcp.xyz/api/mcpAvailable Tools:
search_memories- Search with fuzzy matching and tag filterscreate_memory- Create new memories with tagsupdate_memory- Update existing memoriesdelete_memory- Delete memories by IDlist_tags- List all tags with usage counts
Authentication:
Include your API key in the Authorization header:
Authorization: Bearer YOUR_API_KEYThe API key automatically identifies your user account.
Core Concepts
Understanding the fundamental building blocks of memcp.
Context
Optional metadata field that provides additional context for a memory. Use it to store related information like source, date, or background details that complement the main content.
Content: "User prefers dark mode themes" Context: "From UI/UX feedback session on 2024-01-15" Tags: #preferences #ui #feedback
Content
The main text of your memory. This is the primary information you want to store and retrieve. Content can be up to 10,000 tokens (approximately 40KB of text).
Memories
Individual items that store your information. Each memory has content, optional context, and tags. Memories are searchable and can be organized using tags.
Tags
Labels that organize and categorize your memories. Tags are powerful because they support boolean filtering (AND/OR/NOT operations) and can be combined for precise searches.
Tag examples: #work #personal #urgent #project-alpha #ideas #reference
Boolean Tag Filtering
Combine tags with logical operators for precise filtering:
work AND urgent- Memories with both tagswork OR personal- Memories with either tagwork AND NOT archived- Work memories that aren't archived(work OR personal) AND urgent- Urgent memories from work or personal
Fuzzy Search
memcp uses trigram-based fuzzy search to find memories even with typos or partial matches. Search for "dark mode" and it will find "darkmode", "dark theme", and similar variations.
Recommended System Prompts
Add these prompts to your AI system instructions to enable optimal memory usage.
General Memory Usage
You have access to a memory system through memcp. Use it to: 1. Store important user preferences, decisions, and context 2. Search for relevant past information before making recommendations 3. Update memories when user preferences change 4. Create memories with descriptive tags for easy retrieval When to create memories: - User expresses preferences or requirements - Important decisions are made - Context that might be needed later - Project-specific information When to search memories: - Before making recommendations - When user references past conversations - To maintain consistency across sessions - To recall user preferences and context In general, first look up the available tags for an understanding of what is in the memory system. Then to search for memories, search by tag or keyword. To add memories, try to keep the content as short as possible. Always use the pre-existing tags that are already in the system.
Best Practices
Tag Organization
- Use consistent naming conventions (lowercase, hyphens for spaces)
- Leverage hierarchical tags: #project, #project-alpha, #project-alpha-ui
- Use category tags: #work, #personal, #learning, #reference
Memory Structure
- Keep content focused and specific
- Use context field for metadata and background
- Include relevant relationships in context
- Avoid storing temporary or quickly outdated information
Search Optimization
- Include key terms in content for better fuzzy matching
- Use descriptive tags that capture the essence
- Combine text search with tag filtering for precision
- Regularly update and maintain memory accuracy
MCP Tools Documentation
Complete reference for all available memcp MCP tools.
search_memories
Search memories with fuzzy matching and tag filters
Parameters:
query(string, optional): Text to search for using fuzzy matchingtagFilter(string, optional): Boolean tag filter expressiontags(array, optional): Array of tag names (AND logic)
Example:
search_memories({
query: "dark mode preferences",
tagFilter: "work AND NOT archived"
})create_memory
Create a new memory with tags
Parameters:
content(string, required): Main memory content (max 10,000 tokens)context(string, optional): Additional metadata or backgroundtags(array, optional): Array of tag names
Example:
create_memory({
content: "User prefers dark mode themes",
context: "From UI/UX feedback session",
tags: ["preferences", "ui", "dark-mode"]
})update_memory
Update an existing memory
Parameters:
memoryId(string, required): ID of memory to updatecontent(string, optional): New contentcontext(string, optional): New contexttags(array, optional): New tags array
Example:
update_memory({
memoryId: "abc123",
content: "Updated: User prefers dark mode with high contrast",
tags: ["preferences", "ui", "dark-mode", "accessibility"]
})delete_memory
Delete a memory by ID
Parameters:
memoryId(string, required): ID of memory to delete
Example:
delete_memory({
memoryId: "abc123"
})list_tags
List all tags with usage counts
Parameters:
- None
Example:
list_tags()
// Returns: [{name: "work", count: 15}, {name: "personal", count: 8}, ...]Authentication
All MCP requests must include authentication:
Authorization: Bearer YOUR_API_KEY
The API key automatically identifies your user account and provides access to your memories only.
Power User Features
Unlock the full potential of memcp with these advanced capabilities.
Boolean Tag Filtering Syntax
Combine tags with logical operators for precise filtering:
# Basic AND work AND urgent # Basic OR work OR personal # NOT operator work AND NOT archived # Complex expressions (work OR personal) AND urgent (project-alpha OR project-beta) AND NOT completed # Multiple conditions work AND urgent AND NOT (archived OR completed)
Use parentheses to group operations and control precedence. AND has higher precedence than OR.
Fuzzy Search Capabilities
memcp uses SQLite FTS5 with trigram indexing for powerful fuzzy matching:
- Handles typos and misspellings automatically
- Matches partial words and substrings
- Supports phrase matching with quotes
- Combines with tag filters for precision
- Fast performance even with large datasets
# These searches will find related results: "dark mode" → "darkmode", "dark theme", "dark-mode" "performanc" → "performance", "performing", "performant" "data base" → "database", "data-base"
Bulk Operations and Efficiency
Tips for managing large numbers of memories efficiently:
- Use descriptive tags to avoid duplicate content
- Combine related information into single memories with good context
- Regularly archive old memories with status tags
- Use consistent naming conventions for tags
- Leverage boolean filters to batch-process similar memories
Common Issues and Solutions
Resolve common problems and get answers to frequently asked questions.
Integration Issues
Q: MCP server not connecting in Claude Desktop
A: Check these common issues:
- Verify the configuration file path is correct
- Ensure your API key is valid and not expired
- Check that npx is available in your PATH
- Restart Claude Desktop after configuration changes
- Check the Claude Desktop logs for error messages
Q: OAuth flow failing in web clients
A: Try these solutions:
- Ensure you're using the same API key for both Client ID and Secret
- Check that pop-ups are not blocked in your browser
- Verify you're logged into memcp.xyz in the same browser
- Try generating a new API key and retry the connection
Authentication Problems
Q: API key authentication not working
A: Verify the following:
- API key is copied correctly without extra spaces
- Authorization header format:
Bearer YOUR_API_KEY - API key hasn't been revoked or expired
- You're using the correct endpoint:
https://memcp.xyz/api/mcp
Performance and Limits
Q: Search is slow or returns no results
A: Performance tips:
- Use specific tags to narrow down search scope
- Try shorter search terms for better fuzzy matching
- Check spelling and try variations
- Use boolean filters to combine tags effectively
Q: Memory content limits
A: Current limitations:
- Maximum content size: 10,000 tokens (~40KB text)
- Maximum context size: 1,000 tokens (~4KB text)
- Maximum tags per memory: 50
- Maximum tag length: 100 characters
General FAQ
Q: Can I share memories between users?
A: No. Memories are strictly isolated per user for privacy and security.
Q: How do I export my memories?
A: Use the search_memories tool with appropriate filters, then save the results. We're working on a dedicated export feature.
Q: Can I use memcp without an AI tool?
A: Yes! You can use the web interface at memcp.xyz to create, search, and manage memories directly.
Data Protection and Privacy
How memcp protects your data and ensures privacy.
Data Isolation
Your memories are completely isolated and private:
- No memory sharing between users under any circumstances
- API keys provide access only to your own memories
- Database queries are scoped to your user ID automatically
- Even administrators cannot access other users' memories
Security Considerations
Important security notes:
- All API requests require valid authentication
- HTTPS enforced for all communications
- Rate limiting prevents abuse and attacks
- Input validation prevents injection attacks
- Regular security audits and updates
How memcp Compares to Other Memory Platforms
Simple answers to help you understand memcp's unique approach.
Q: How does memcp.xyz compare to mem0, Supermemory, and Letta?
A: memcp.xyz is focused on being incredibly simple and easy to use. Most tools on the market are for developers, seeking to build memory systems within their own AI applications and products. memcp.xyz is intended for regular users of AI tools with a fast and powerful Web-UI and integrate easily with a hosted MCP server.