LangChain Agent MCP Server

# Troubleshooting Guide Common issues and solutions for the LangChain Agent MCP Server. ## Local Development Issues ### Python Not Found **Issue:** `python` or `py` command not recognized **Solution:** - Windows: Use `py` instead of `python` - Verify installation: `py --version` - Add Python to PATH if needed ### pip Not Found **Issue:** `pip` command not recognized **Solution:** ```powershell # Windows py -m pip install -r requirements.txt # Linux/Mac python -m pip install -r requirements.txt ``` ### Module Import Errors **Issue:** `ImportError: cannot import name 'Tool'` **Solution:** - Check LangChain version: `pip show langchain` - Update requirements: `pip install -r requirements.txt --upgrade` - Verify imports work: `py -c "from langchain_core.tools import Tool; print('OK')"` ### API Key Not Detected **Issue:** Server warns about missing API key **Solution:** 1. Verify `.env` file exists in project root 2. Check API key format: `OPENAI_API_KEY=sk-proj-...` 3. Restart server after updating `.env` ## Deployment Issues ### Docker Build Fails **Issue:** Build errors or missing dependencies **Solution:** - Check Docker is running: `docker ps` - Verify Dockerfile syntax - Check requirements.txt for version conflicts ### Cloud Run Deployment Fails **Issue:** Container fails to start **Solution:** 1. Check logs: ```powershell gcloud run services logs read langchain-agent-mcp-server --region us-central1 ``` 2. Verify PORT environment variable is handled correctly 3. Check startup script permissions: `chmod +x src/start.sh` ### 401 Unauthorized from OpenAI **Issue:** Agent returns 401 errors **Solution:** 1. Verify API key is set: ```powershell gcloud run services describe langchain-agent-mcp-server --region us-central1 ``` 2. Check key is valid at: https://platform.openai.com/account/api-keys 3. Verify billing/credits on OpenAI account ### Service Returns 404 **Issue:** Endpoints return 404 Not Found **Solution:** - Verify service URL is correct (no typos) - Check service is running: `/health` endpoint - Verify endpoint paths: `/mcp/manifest`, `/mcp/invoke` ## Performance Issues ### Slow Response Times **Issue:** Agent takes too long to respond **Solution:** - Reduce `MAX_ITERATIONS` environment variable - Use faster model: `OPENAI_MODEL=gpt-4o-mini` - Increase Cloud Run resources: `--memory 4Gi --cpu 4` ### Agent Hits Iteration Limit **Issue:** "Agent stopped due to iteration limit" **Solution:** - Increase `MAX_ITERATIONS` (default: 10) - Simplify queries - Check if agent is stuck in a loop ## Integration Issues ### SlashMCP 404 Errors **Issue:** MCP gateway returns 404 **Solution:** 1. Verify service URL is correct (check for typos) 2. Test endpoints directly: `/health`, `/mcp/manifest` 3. Check CORS configuration 4. Verify service is publicly accessible ### MCP Client Connection Issues **Issue:** Cannot connect to server **Solution:** - Verify service is running: Check `/health` - Check network connectivity - Verify URL format: `https://...` not `http://...` - Check firewall/security settings ## Getting Help ### Check Logs **Local:** ```powershell # Server logs are printed to console ``` **Cloud Run:** ```powershell gcloud run services logs read langchain-agent-mcp-server \ --platform managed \ --region us-central1 \ --limit 50 ``` ### Verify Configuration ```powershell # Check environment variables gcloud run services describe langchain-agent-mcp-server \ --region us-central1 \ --format="value(spec.template.spec.containers[0].env)" ``` ### Test Endpoints ```powershell # Health Invoke-WebRequest -Uri "https://langchain-agent-mcp-server-554655392699.us-central1.run.app/health" # Manifest Invoke-WebRequest -Uri "https://langchain-agent-mcp-server-554655392699.us-central1.run.app/mcp/manifest" ``` ## Common Error Messages ### "OPENAI_API_KEY not set" - Set in `.env` file for local development - Set in Cloud Run environment variables for deployment ### "Cannot import name 'Tool'" - Update LangChain: `pip install langchain-core --upgrade` - Check version compatibility ### "Container failed to start" - Check Cloud Run logs - Verify Docker image builds successfully - Check PORT environment variable handling ### "MCP gateway request failed with status 404" - Verify service URL is correct - Check service is running - Test endpoints directly --- **Still having issues?** Check the [GitHub Issues](https://github.com/mcpmessenger/LangchainMCP/issues) or create a new one.