Guides · ChatGPT
Configure ChatGPT Desktop with MCP servers for AI-assisted workflows, knowledge retrieval, and task automation.
ChatGPT Desktop (macOS and Windows) supports MCP (Model Context Protocol) as of early 2025. This allows ChatGPT to access external tools and data sources, including AACFlow's workflows, knowledge bases, and connectors. You need ChatGPT Plus or Pro subscription for MCP features.
1. Download ChatGPT Desktop from chat.openai.com/download (macOS 11+ or Windows 10+). 2. Sign in with your ChatGPT Plus/Pro account. 3. Enable Developer Mode in Settings (required for MCP). 4. Verify Node.js 18+ is installed: node --version
1. Create or edit the MCP config file: macOS: ~/Library/Application Support/ChatGPT/chatgpt_mcp_config.json Windows: %APPDATA%\OpenAI\ChatGPT\chatgpt_mcp_config.json 2. Add the AACFlow server: { "mcpServers": { "aacflow": { "command": "npx", "args": ["-y", "@aacflow/mcp-server"], "env": { "AACFLOW_API_KEY": "ak_your_key" } } } } 3. Restart ChatGPT Desktop. Look for the MCP indicator (hammer icon) in the chat interface.
1. Open ChatGPT Desktop and start a new conversation. 2. Ask: "Search my AACFlow knowledge base for documentation about the API authentication flow." 3. ChatGPT will use the MCP server to query your AACFlow workspace. 4. Extend with: "Now create a workflow that sends a Slack notification when new docs are added." 5. ChatGPT can create, test, and deploy the workflow through AACFlow's MCP tools.
• Never point filesystem MCP at your home directory — use a specific project folder. • Store API keys in environment variables, not hardcoded in config files. • Use the companion window (Alt+Space) for quick access while coding. • Combine MCP servers: AACFlow for workflows + Brave Search for real-time data. • For enterprise: deploy ChatGPT via winget or Intune with pre-configured MCP settings. • Test MCP connectivity by asking ChatGPT to list your AACFlow projects.
