MemoryGraph

# MemoryGraph Setup for ChatGPT Desktop Get persistent memory working in ChatGPT Desktop in under 2 minutes. ## Prerequisites - ChatGPT Desktop app (latest version with MCP support) - Python 3.10+ - pipx installed (`pip install --user pipx && pipx ensurepath`) ## MCP Support Status As of 2025, OpenAI has officially adopted the Model Context Protocol (MCP) and integrated it across their products: - **ChatGPT Desktop**: Full MCP client support (March 2025+) - **Developer Mode**: Beta feature with complete MCP tool support - **OpenAI Agents SDK**: Native MCP integration - **Responses API**: MCP-compatible ChatGPT Desktop now includes built-in MCP client capabilities for seamless integration with external tools, real-time data sources, and specialized workflows. ## Choose Your Backend MemoryGraph supports two backend options: | Feature | Local (SQLite) | Cloud | |---------|---------------|-------| | **Setup** | Zero-config | API key required | | **Data Location** | `~/.memorygraph/` | memorygraph.dev | | **Multi-device** | No | Yes | | **Team sharing** | No | Yes | | **Offline** | Yes | No | | **Cost** | Free | Free tier available | **New users**: We recommend starting with **Cloud** for multi-device sync, or **Local** for single-machine use. --- ## Quick Start (Local Backend) ### 1. Install MemoryGraph ```bash pipx install memorygraphMCP ``` Verify installation: ```bash memorygraph --version ``` ### 2. Configure ChatGPT Desktop The exact configuration method depends on your ChatGPT Desktop version: #### Option A: Via Developer Mode (Beta) 1. Open ChatGPT Desktop 2. Enable Developer Mode in Settings 3. Navigate to **Settings > MCP Servers** 4. Click **Add Server** 5. Enter: - Name: `memorygraph` - Command: `memorygraph` - Args: (leave empty or add `--profile extended` for more tools) #### Option B: Via Configuration File ChatGPT Desktop may use a configuration file similar to Claude Desktop. Check for: **macOS**: ```bash ~/Library/Application\ Support/ChatGPT/chatgpt_config.json ``` **Windows**: ``` %APPDATA%\ChatGPT\chatgpt_config.json ``` Add the MCP server configuration: ```json { "mcpServers": { "memorygraph": { "command": "memorygraph", "args": [] } } } ``` **Note**: Configuration file location may vary by version. Check ChatGPT Desktop documentation for your specific version. ### 3. Handle PATH Issues (if needed) If ChatGPT Desktop cannot find the `memorygraph` command: **Option A: Use full path** (recommended) ```bash # Find your installation path which memorygraph # Use full path in config { "command": "/Users/yourname/.local/bin/memorygraph" } ``` **Option B: Create symlink** ```bash # Create symlink to system directory sudo ln -s ~/.local/bin/memorygraph /usr/local/bin/memorygraph # Then use simple command { "command": "memorygraph" } ``` ### 4. Restart ChatGPT Desktop Completely quit and restart ChatGPT Desktop to load the MCP server. ### 5. Verify Connection In a new conversation, ask: ``` What memory tools do you have available? ``` You should see MemoryGraph tools like `store_memory`, `recall_memories`, `search_memories`, etc. ## First Memory Try storing your first memory: ``` Store this for later: This project uses TypeScript with strict mode enabled ``` Then retrieve it: ``` What do you know about this project's TypeScript setup? ``` --- ## Quick Start (Cloud Backend) Cloud backend syncs memories across all your devices and enables team collaboration. ### 1. Get Your API Key 1. Sign up at [app.memorygraph.dev](https://app.memorygraph.dev) 2. Copy your API key (starts with `mg_`) ### 2. Install MemoryGraph ```bash pipx install memorygraphMCP ``` ### 3. Configure ChatGPT Desktop with Cloud Backend Edit your configuration file: **macOS**: `~/Library/Application Support/ChatGPT/chatgpt_config.json` **Windows**: `%APPDATA%\ChatGPT\chatgpt_config.json` ```json { "mcpServers": { "memorygraph": { "command": "/Users/yourname/.local/bin/memorygraph", "args": ["--backend", "cloud"], "env": { "MEMORYGRAPH_API_KEY": "mg_your_api_key_here" } } } } ``` **Note**: Use the full path from `which memorygraph`. ### 4. Restart ChatGPT Desktop and Verify 1. Completely quit ChatGPT Desktop 2. Reopen and ask: "What memory tools do you have available?" --- ## Migrating from Local to Cloud Already using local SQLite and want to switch to cloud? ### Step 1: Export Local Memories ```bash memorygraph export --output memories-backup.json ``` ### Step 2: Import to Cloud ```bash export MEMORYGRAPH_API_KEY=mg_your_key_here memorygraph import --backend cloud --input memories-backup.json ``` ### Step 3: Update ChatGPT Desktop Configuration Update your config file: ```json { "mcpServers": { "memorygraph": { "command": "/Users/yourname/.local/bin/memorygraph", "args": ["--backend", "cloud"], "env": { "MEMORYGRAPH_API_KEY": "mg_your_api_key_here" } } } } ``` ### Step 4: Restart ChatGPT Desktop Quit completely and reopen to apply the new configuration. See [CLOUD_BACKEND.md](../CLOUD_BACKEND.md) for detailed migration options. --- ## Configuration Options ### Extended Mode (More Tools) ```json { "mcpServers": { "memorygraph": { "command": "memorygraph", "args": ["--profile", "extended"] } } } ``` Adds statistics and complex query tools (11 total tools). ### Custom Database Location ```json { "mcpServers": { "memorygraph": { "command": "memorygraph", "args": [], "env": { "MEMORY_SQLITE_PATH": "/path/to/your/memory.db" } } } } ``` ### Extended Mode with Neo4j ```json { "mcpServers": { "memorygraph": { "command": "memorygraph", "args": ["--profile", "extended", "--backend", "neo4j"], "env": { "MEMORY_NEO4J_URI": "bolt://localhost:7687", "MEMORY_NEO4J_USER": "neo4j", "MEMORY_NEO4J_PASSWORD": "your-password" } } } } ``` ## Troubleshooting ### Server Not Connecting 1. **Verify memorygraph is installed**: ```bash which memorygraph memorygraph --version ``` 2. **Check Developer Mode is enabled** (if using beta): - Settings > Developer Mode > ON - MCP Servers section should be visible 3. **Test server manually**: ```bash echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0.0"}},"id":1}' | memorygraph ``` 4. **Use full path in config**: ```json { "command": "/Users/yourname/.local/bin/memorygraph" } ``` ### Tools Not Appearing 1. **Restart completely**: Quit ChatGPT Desktop (not just close window) and reopen 2. **Check MCP server status**: Look for server status indicator in Settings 3. **Verify JSON syntax**: Validate your config file at https://jsonlint.com 4. **Check logs**: Look for ChatGPT Desktop logs in: - macOS: `~/Library/Logs/ChatGPT/` - Windows: `%APPDATA%\ChatGPT\logs\` ### Command Not Found Error Similar to Claude Desktop, ChatGPT Desktop may have restricted PATH: ```bash # Solution 1: Use full path which memorygraph # Copy this path # Add to config: "/Users/yourname/.local/bin/memorygraph" # Solution 2: Create symlink sudo ln -s ~/.local/bin/memorygraph /usr/local/bin/memorygraph ``` ### Permission Issues On macOS, you may need to grant ChatGPT Desktop permissions: 1. System Settings > Privacy & Security > Full Disk Access 2. Add ChatGPT Desktop to the list 3. Restart ChatGPT Desktop ## Tips for ChatGPT Desktop Users 1. **Use Developer Mode for full MCP access** - Regular mode may have limited tool support 2. **Store context between sessions** - Build up project knowledge over time 3. **Query before starting** - Check if you've solved similar problems 4. **Save successful patterns** - Remember what works for future reference 5. **Link related memories** - Create relationships between problems and solutions ## Recommended: Memory Protocol For consistent memory usage, configure ChatGPT with memory guidelines. If ChatGPT Desktop supports custom instructions, add: ```markdown ## Memory Protocol ### REQUIRED: Before Starting Work You MUST use `recall_memories` before any task. Query by project, tech, or task type. ### REQUIRED: Automatic Storage Triggers Store memories on ANY of: - **Git commit** → what was fixed/added - **Bug fix** → problem + solution - **Version release** → summarize changes - **Architecture decision** → choice + rationale - **Pattern discovered** → reusable approach ### Timing Mode (default: on-commit) `memory_mode: immediate | on-commit | session-end` ### Memory Fields - **Type**: solution | problem | code_pattern | fix | error | workflow - **Title**: Specific, searchable (not generic) - **Content**: Accomplishment, decisions, patterns - **Tags**: project, tech, category (REQUIRED) - **Importance**: 0.8+ critical, 0.5-0.7 standard, 0.3-0.4 minor - **Relationships**: Link related memories when they exist Do NOT wait to be asked. Memory storage is automatic. ``` ## OpenAI Apps SDK Integration ChatGPT Desktop uses the OpenAI Apps SDK, which has MCP as its backbone. This enables: - **Rich interactive applications** inside ChatGPT - **Standardized tool integration** via MCP protocol - **Bidirectional communication** between tools and ChatGPT - **Resource discovery** through MCP servers MemoryGraph leverages this SDK integration to provide persistent memory capabilities. ## Features Available via MCP ChatGPT Desktop MCP integration provides: - **Extended capabilities**: Access specialized tools directly in conversations - **Real-time information**: Retrieve data beyond ChatGPT's training cutoff - **Workflow automation**: Connect ChatGPT to existing tools and systems - **Tool chaining**: Combine multiple MCP tools for complex workflows - **Persistent context**: Store and recall information across sessions ## Version Compatibility - **MCP Protocol**: 2024-11-05 specification - **ChatGPT Desktop**: Version with MCP support (March 2025+) - **MemoryGraph**: v0.9.0+ recommended for full compatibility Check for updates regularly as OpenAI continues to enhance MCP integration. ## Example Full Configuration Here's a complete working example (macOS): ```json { "mcpServers": { "memorygraph": { "command": "/Users/john/.local/bin/memorygraph", "args": ["--profile", "extended"], "env": { "MEMORY_SQLITE_PATH": "/Users/john/.memorygraph/memory.db", "MEMORY_LOG_LEVEL": "INFO" } } } } ``` Remember to: 1. Replace `/Users/john` with your actual home directory path 2. Use the path from `which memorygraph` 3. Ensure valid JSON syntax (commas, quotes, braces) ## Next Steps - [Full Configuration Reference](../CONFIGURATION.md) - [All Tool Profiles](../TOOL_PROFILES.md) - [Troubleshooting Guide](../TROUBLESHOOTING.md) - [OpenAI Documentation](https://platform.openai.com/) ## Resources - [OpenAI MCP Announcement](https://platform.openai.com/docs/guides/mcp) - [MCP Protocol Specification](https://modelcontextprotocol.io) - [ChatGPT Desktop Download](https://openai.com/chatgpt/desktop) --- **Works with**: ChatGPT Desktop (macOS, Windows) with MCP support **Transport**: stdio **Profiles**: core (9 tools), extended (11 tools) **Status**: Official MCP support as of March 2025 **Note**: ChatGPT Desktop's MCP implementation is evolving. Check OpenAI's documentation for the latest configuration details specific to your version.