Connect Ditto to Any AI Assistant
Your Ditto memories are now accessible to any AI assistant that supports the Model Context Protocol (MCP). This means Claude, ChatGPT, Cursor, Antigravity, and dozens of other tools can tap into your personal knowledge graph—giving them context about your projects, preferences, and work history.
In this guide, we’ll show you exactly how to connect Ditto to your favorite AI tools.
What is MCP?
The Model Context Protocol is an open standard that lets AI assistants access external context sources. Think of it as a universal adapter that connects LLMs to your data.
Ditto implements MCP with:
- OAuth 2.1 for seamless, secure authentication
- SSE (Server-Sent Events) for real-time streaming
- Rich memory tools for searching, exploring, and retrieving your knowledge graph
What Can AI Assistants Access?
When you connect Ditto via MCP, AI assistants get access to powerful memory tools:
🔍 Search Tools
search_subjects- Find topics in your knowledge graph by semantic similaritysearch_memories- Search all memories with learned retrieval weightssearch_memories_in_subject- Get memories linked to specific subjects
📊 Resources
memory://profile- Your profile, personality summary, and preferencesmemory://stats- Memory statistics and top subjectsmemory://capabilities- Available tools and API limits
🧠 Memory Network
get_memory_network- Explore related memories and their connectionsget_memory_by_pair_id- Retrieve specific memory pairs
This means your AI assistant can:
- Remember past conversations and projects
- Understand your work style and preferences
- Find relevant context from your entire knowledge base
- Make connections across different topics you’ve discussed
Quick Start: Choose Your Setup Method
There are two ways to connect Ditto to AI assistants:
- OAuth (Recommended) - One-click setup, no API keys needed
- API Keys - Manual setup for tools that don’t support OAuth
Method 1: OAuth Setup (Claude.ai)
The easiest way to connect. No API keys, no configuration files.
Steps:
- Go to Claude.ai → Settings → Connectors
- Click “Add custom connector”
- Enter MCP URL:
https://api.heyditto.ai/mcp/sse - Click “Connect” and sign in with your Ditto account
- Approve the authorization
That’s it! Claude now has access to your Ditto memories.
Method 2: API Key Setup
For tools that require manual configuration (Cursor, Claude Code, etc.), you’ll need an API key.
Get Your API Key:
- Log in to Ditto
- Go to Settings → MCP API Keys
- Click “Create Key” and give it a name
- Copy the key immediately - you won’t see it again!
Now follow the setup guide for your specific tool below.
Integration Guides by Platform
🤖 Claude Code
Claude Code supports MCP via a simple terminal command:
claude mcp add --transport sse ditto https://api.heyditto.ai/mcp/sse \
--header "Authorization: Bearer YOUR_API_KEY"
Replace YOUR_API_KEY with the key you created above.
Usage: Open any project in Claude Code and your Ditto memories will automatically be available. Claude can search your knowledge graph and retrieve relevant context as you work.
✏️ Cursor IDE
Cursor uses a JSON configuration file for MCP servers.
Setup:
- Open
~/.cursor/mcp.json(create it if it doesn’t exist) - Add this configuration:
{
"mcpServers": {
"ditto": {
"url": "https://api.heyditto.ai/mcp/sse",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
- Restart Cursor
Usage: When you chat with Cursor’s AI, it can now search your Ditto memories to find relevant context. Try asking it to “search my memories for [topic]” or reference past projects.
🖥️ Claude Desktop
Claude Desktop works the same way as Cursor:
Setup:
- Locate your Claude Desktop config:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
- macOS:
- Add the Ditto MCP server:
{
"mcpServers": {
"ditto": {
"url": "https://api.heyditto.ai/mcp/sse",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
- Restart Claude Desktop
🚀 Antigravity IDE
Antigravity uses a slightly different config format (note serverUrl instead of url):
Setup:
- Open
~/.gemini/antigravity/mcp_config.json - Add this configuration:
{
"mcpServers": {
"ditto": {
"serverUrl": "https://api.heyditto.ai/mcp/sse",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
- Restart Antigravity
💬 ChatGPT (via Custom GPT)
While ChatGPT doesn’t natively support MCP, you can create a Custom GPT that connects to Ditto’s API:
Setup:
- Go to ChatGPT → Explore GPTs → Create
- In “Configure” → “Actions”, import this OpenAPI schema:
openapi: 3.0.0
info:
title: Ditto Memory API
version: 1.0.0
servers:
- url: https://api.heyditto.ai
paths:
/mcp:
post:
summary: Access Ditto memories via MCP
security:
- BearerAuth: []
requestBody:
content:
application/json:
schema:
type: object
responses:
"200":
description: MCP response
components:
securitySchemes:
BearerAuth:
type: http
scheme: bearer
- Add your API key in the “Authentication” section
- Save and use your custom GPT
Note: This provides basic access. For full MCP functionality, use tools that natively support the protocol.
🛠️ Other MCP-Compatible Tools
The configuration pattern is consistent across most MCP clients:
{
"mcpServers": {
"ditto": {
"url": "https://api.heyditto.ai/mcp/sse",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
Or via command line (if supported):
[tool] mcp add --transport sse ditto https://api.heyditto.ai/mcp/sse \
--header "Authorization: Bearer YOUR_API_KEY"
Check your tool’s MCP documentation for the exact config file location.
Using Ditto in AI Conversations
Once connected, your AI assistant can access your memories. Here are some powerful ways to use it:
Example Prompts
Context Retrieval:
Search my Ditto memories for information about the React migration project
Project Continuity:
What were the key decisions we made last week about the API redesign?
(AI will search your memories for relevant context)
Personalized Responses:
Based on my work style and preferences, suggest the best approach for this feature
Subject Exploration:
Show me all my memories related to "performance optimization"
The AI will automatically use Ditto’s MCP tools (search_memories, search_subjects, etc.) to find relevant information from your knowledge graph.
Security & Privacy
API Key Best Practices
- Never commit API keys to git - they provide full access to your memories
- Create separate keys per tool - easier to revoke if needed
- Revoke unused keys - clean up old keys in Settings → MCP API Keys
- Monitor usage - check “Last Used” timestamps in your API key list
OAuth Security
OAuth is more secure than API keys because:
- No credential storage in config files
- Automatic token rotation
- Fine-grained permission scopes
- Easy revocation from Ditto settings
Use OAuth whenever your AI tool supports it.
What Data is Shared?
When an AI assistant connects to Ditto via MCP:
- ✅ It can search and read your memories
- ✅ It can access your profile and preferences
- ❌ It cannot create, edit, or delete memories
- ❌ It cannot access other users’ data
All MCP access is read-only and scoped to your user account.
Troubleshooting
”Not Authenticated” Error
Problem: AI tool can’t connect to Ditto
Solutions:
- Verify your API key is correct (no extra spaces)
- Check that the key hasn’t been revoked
- Ensure the Authorization header format is:
Bearer YOUR_API_KEY - Restart your AI tool after adding the configuration
”No Results” When Searching
Problem: AI searches return empty results
Solutions:
- Make sure you have memories in Ditto (chat with Ditto first!)
- Try broader search terms
- Check the
minSimilaritythreshold (lower = more results) - Verify the AI is actually calling MCP tools (check tool invocations)
Config File Not Found
Problem: Can’t locate the MCP config file
Solutions:
- Create the file manually at the path specified
- Ensure the parent directories exist
- Check tool-specific documentation for the exact path
- On Windows, use
%APPDATA%instead of~/
OAuth Flow Fails
Problem: Authorization doesn’t complete
Solutions:
- Make sure you’re logged into Ditto first
- Check that pop-ups aren’t blocked
- Try in an incognito/private window
- Clear browser cache and try again
Advanced: Using MCP Tools Directly
For developers, you can also call Ditto’s MCP tools programmatically:
// Example: Search subjects
{
"method": "tools/call",
"params": {
"name": "search_subjects",
"arguments": {
"query": "machine learning projects",
"topK": 10,
"minSimilarity": 0.1
}
}
}
// Example: Search memories
{
"method": "tools/call",
"params": {
"name": "search_memories",
"arguments": {
"queries": ["API design", "authentication"],
"stripImages": true
}
}
}
See the MCP specification for the full protocol details.
What’s Next?
We’re constantly expanding Ditto’s MCP capabilities. Coming soon:
- Write access - AI assistants that can create memories
- Real-time sync - Live updates when memories change
- Custom tools - Define your own MCP tools for specific workflows
- Analytics - Track which memories your AI uses most
Get Started Today
Ready to give your AI assistants access to your knowledge?
- Sign up for Ditto if you haven’t already
- Create an MCP API key in Settings
- Follow the integration guide for your favorite AI tool
- Start having conversations with full context
Have questions? Join our Discord community or email us at hello@heyditto.ai.
Built with the Model Context Protocol (MCP) - the open standard for AI context. Learn more at modelcontextprotocol.io