Working with Logs
The Logs page is where you investigate what happened during Flow and Agent executions. Every execution produces a log entry with a step-by-step breakdown you can drill into.
Viewing execution details
Click any row in the Logs list to open the execution detail view:
- Timeline — each step in execution order with status indicators and duration.
- Step inspector — select a step to see its full input, output, and configuration.
- Token usage — for prompt steps, the number of input and output tokens consumed.
- Context window — for agent turns, a segmented bar splitting the estimated input tokens across the system prompt, tool definitions, record context, conversation history, tool results, and current message, plus overall utilization against the model’s context window (for example, “140.3k of 200k (70%)”). These values are estimates and can differ from your billed tokens.
- Trace: a tree of the execution’s iterations, tool calls, and model turns, so you can follow the agent loop end to end. When context estimates are available, a “Context per iteration” chart sits above the tree. It uses the same categories and colors as the context window bar, with one stacked column per iteration, so you can see how context grows as the run progresses. A dashed line marks the context window ceiling when the model’s context size is known.
Agent executions
Agent logs show each iteration of the agent loop, including tool calls, tool results, and the agent’s reasoning at each turn. This helps you understand why the agent chose specific tools and how it arrived at its final response.
Add a run to an eval
Open an agent execution and click Add to eval to turn the run into a test case. Pick the target eval (or let Runtype create a default one for the agent), choose which recorded tool call the eval should grade as the next step, and save. The eval freezes the conversation up to that fork point and grades what the agent does next. After saving you see whether the case is fully replayable, or only partially (when a recorded tool output was too large to capture in full). See Managing eval suites for how the captured case runs.
Searching and filtering
Use search and filters to find specific executions in your log history.
Filtering options
- Flow or Agent — narrow to executions of a specific Flow or Agent.
- Status — show only successful, failed, or in-progress executions.
- Date range — limit results to a specific time window.
Search
Use the search bar to find executions by keywords in the execution data, including Record metadata and step outputs.
Live mode
Toggle Live in the Logs view to see new executions appear as they start and complete. This is useful when testing a Flow or monitoring a scheduled batch run.
Live mode automatically pauses when you scroll up to review older entries and resumes when you scroll back to the bottom.
While live mode is on, a context strip above the list shows up to three pills for the most recently active executions, each estimating how close that run is to the model’s context limit. Click a pill to open that execution’s details.
Exporting logs
Use the export action to download log data as JSON or CSV. Exports include the execution metadata, step details, and timing information for each entry in your current filtered view.
Debugging failed executions
When a Flow or Agent execution fails, logs show you exactly which step caused the failure and why.
Identifying the failed step
Open a failed execution and look for the step marked with an error status in the timeline. The step inspector shows the error message and the input that caused the failure.
Common failure patterns
Re-running after a fix
After fixing the issue, re-run the Flow manually from the Flows page or wait for the next scheduled execution if the Flow is on a Schedule.
Next steps
- Debugging flows for step-level troubleshooting
- Flow execution failures for common error patterns
- What are Logs? for a high-level overview