FCCS MCP Agentic Server

# ChatGPT Custom GPT Configuration Guide This guide explains how to configure ChatGPT (Custom GPT) to use the FCCS MCP Server. ## Option 1: Using Cloud Run (Recommended) ### Step 1: Deploy to Cloud Run First, deploy the server to Google Cloud Run (see `QUICK_DEPLOY.md`): ```powershell .\deploy-cloud-run.ps1 ``` After deployment, you'll get a URL like: ``` https://fccs-mcp-server-xxxxx-uc.a.run.app ``` ### Step 2: Configure ChatGPT Custom GPT 1. **Go to ChatGPT Custom GPTs:** - Visit: https://chat.openai.com/gpts - Click "Create" or "Edit" on your GPT 2. **Configure the Action:** - Go to the "Actions" section - Click "Create new action" - Choose "API" or "OpenAPI schema" 3. **Add API Configuration:** **Schema URL (if using OpenAPI):** ``` https://your-cloud-run-url.run.app/openapi.json ``` **Or Manual Configuration:** - **Base URL:** `https://your-cloud-run-url.run.app` - **Authentication:** None (or API Key if you enable auth) 4. **Available Endpoints:** - `GET /health` - Health check - `GET /tools` - List all available tools - `POST /execute` - Execute any tool - `POST /message` - MCP-compatible endpoint ### Step 3: Test Connection In ChatGPT, try: ``` "List all available FCCS tools" ``` ChatGPT should call `/tools` endpoint and show you all available tools. ## Option 2: Using Local Server ### Step 1: Start Local Server ```powershell # Set environment variables $env:FCCS_URL = "your-fccs-url" $env:FCCS_USERNAME = "your-username" $env:FCCS_PASSWORD = "your-password" # Or use mock mode $env:FCCS_MOCK_MODE = "true" # Start server python -m web.server ``` Server will run on: `http://localhost:8080` ### Step 2: Expose Local Server (ngrok) Since ChatGPT needs a public URL, use ngrok: 1. **Install ngrok:** ```powershell # Download from https://ngrok.com/download # Or use chocolatey choco install ngrok ``` 2. **Start ngrok tunnel:** ```powershell ngrok http 8080 ``` 3. **Copy the public URL:** ``` https://xxxx-xx-xx-xx-xx.ngrok-free.app ``` 4. **Use this URL in ChatGPT Custom GPT configuration** ## Option 3: Direct API Integration ### Using the `/message` Endpoint (MCP-compatible) ChatGPT can call the MCP-compatible endpoint: **Request Format:** ```json POST /message { "method": "tools/list", "params": {} } ``` **Or:** ```json POST /message { "method": "tools/call", "params": { "name": "get_application_info", "arguments": {} } } ``` ### Using Standard REST Endpoints **List Tools:** ```json GET /tools ``` **Execute Tool:** ```json POST /execute { "tool_name": "get_application_info", "arguments": {}, "session_id": "chatgpt-session" } ``` ## ChatGPT Custom GPT Instructions Add these instructions to your Custom GPT: ``` You are an Oracle FCCS (Financial Consolidation and Close) assistant. You have access to the FCCS API through the configured actions. Available capabilities: - Get application information and API versions - List and execute business rules and data rules - Query dimensions and members - Manage consolidation journals - Export and import data slices - Generate reports - Run consolidation processes When the user asks about FCCS data or operations, use the available tools/actions. Always provide clear, structured responses with the data retrieved. For financial analysis: - Use smart_retrieve for quick data queries - Use export_data_slice for complex grid exports - Format financial numbers with proper currency symbols and commas Be helpful and explain what operations you're performing. ``` ## Example Prompts for ChatGPT Once configured, you can use prompts like: 1. **"Get information about the FCCS application"** - Calls: `get_application_info` 2. **"Show me the top 10 performing entities in 2024"** - Calls: `smart_retrieve` with Net Income data 3. **"List all dimensions in the application"** - Calls: `get_dimensions` 4. **"Get members of the Entity dimension"** - Calls: `get_members` with dimension_name="Entity" 5. **"Run the consolidation business rule"** - Calls: `run_business_rule` 6. **"Show me journals for January 2025"** - Calls: `get_journals` with filters ## Authentication (Optional) If you want to secure your API: ### Option 1: API Key Authentication 1. **Update web server** to require API key 2. **Set API key in ChatGPT Custom GPT** configuration 3. **Add header:** `Authorization: Bearer YOUR_API_KEY` ### Option 2: Cloud Run Authentication 1. **Remove `--allow-unauthenticated`** from deployment 2. **Enable authentication** in Cloud Run 3. **Configure service account** in ChatGPT ## Troubleshooting ### ChatGPT can't connect - Verify the server URL is accessible - Check Cloud Run service is running - Test with: `curl https://your-url.run.app/health` ### Tools not showing - Check `/tools` endpoint returns data - Verify JSON format is correct - Check ChatGPT action configuration ### Authentication errors - Verify API key is set correctly - Check Cloud Run IAM permissions - Ensure service account has access ### Timeout errors - Increase Cloud Run timeout: `--timeout 600` - Check FCCS API response times - Consider async operations for long-running tasks ## Testing the Configuration 1. **Test health endpoint:** ```powershell curl https://your-url.run.app/health ``` 2. **Test tools list:** ```powershell curl https://your-url.run.app/tools ``` 3. **Test tool execution:** ```powershell curl -X POST https://your-url.run.app/execute ` -H "Content-Type: application/json" ` -d '{"tool_name": "get_application_info", "arguments": {}, "session_id": "test"}' ``` ## Next Steps 1. Deploy to Cloud Run (see `QUICK_DEPLOY.md`) 2. Configure ChatGPT Custom GPT with the Cloud Run URL 3. Test with simple queries 4. Add custom instructions for your use case 5. Set up monitoring and logging ## Support For issues: - Check Cloud Run logs: `gcloud run logs tail fccs-mcp-server --region us-central1` - Test endpoints directly with curl/Postman - Verify environment variables are set correctly