Agents And Workflows
Debugging Workflows
Use the Trace Viewer to visualize agent cognitive paths and execution errors.
Execution Logs
Click on any failed workflow run to open the Trace Viewer. You can inspect the exact prompt injected into the LLM and the raw JSON response returned by the model.
Common Errors
Most ToolExecutionError faults arise from strictly typed schema mismatches. Verify your MCP tool definitions against the connected API spec.