Token consumption
Total input and output tokens consumed, broken down by time period. Track daily, weekly, and monthly consumption trends to forecast credit usage.
Logs & Metrics
Dashboard
Analytics
The Logs & Metrics page provides a detailed, real-time view of everything happening across your organization’s agents, tools, and LLM usage. Use it to monitor live runs, investigate failures, and understand resource consumption.
Live agent runs
Section titled “Live agent runs”The top of the page displays a real-time feed of currently running agent executions. Each entry shows:
| Field | Description |
|---|---|
| Run ID | Unique identifier for the agent execution |
| Agent | The agent handling the run |
| Status | Current state: running, completed, failed, or timed out |
| Started | Timestamp when the run began |
| Duration | Elapsed time since the run started (updates live for active runs) |
| Tool calls | Number of tool invocations made so far in this run |
| Tokens used | Total LLM tokens consumed by the run |
Click any row to expand the full execution trace, including each tool call, its input/output, and the LLM reasoning steps.
Tool execution metrics
Section titled “Tool execution metrics”The tool metrics section aggregates data across all tool invocations in your organization:
A breakdown of which tools are being called most frequently. Use this to understand which tools are critical to your workflows and which may be underutilized.
Key data points:
- Total calls — aggregate count over the selected time range
- Calls per tool — ranked list of tools by invocation count
- Calls per agent — which agents are driving the most tool usage
Track the reliability of your tool executions:
- Success rate — percentage of tool calls that returned a valid result
- Failure rate — percentage that returned an error or timed out
- Top failing tools — tools with the highest error counts, sorted by frequency
- Error categories — common failure reasons (timeout, connector unavailable, validation error, internal error)
Understand how long tool executions take:
- Average execution time — mean duration across all tool calls
- P50 / P95 / P99 latency — percentile breakdowns to identify outliers
- Slowest tools — ranked by average execution time
- Latency trends — graph showing execution time over the selected time range
Agent run statistics
Section titled “Agent run statistics”Aggregate statistics for all agent runs across your organization:
| Metric | Description |
|---|---|
| Total runs | Number of agent executions in the selected time range |
| Average duration | Mean time from run start to completion |
| Completion rate | Percentage of runs that finished without errors |
| Timeout rate | Percentage of runs that exceeded the maximum execution time |
| Average tool calls per run | Mean number of tool invocations per agent execution |
| Average tokens per run | Mean LLM token consumption per execution |
LLM usage metrics
Section titled “LLM usage metrics”Monitor your organization’s LLM consumption in detail:
Model usage breakdown
See which LLM models are being used and their relative share of total token consumption. Useful for understanding cost distribution across model tiers.
Cost per conversation
Average credit cost per conversation session, broken down by agent. Identify which agents are the most and least cost-efficient.
Peak usage periods
Hourly and daily heatmaps showing when LLM usage is highest. Use this data to anticipate capacity needs and optimize credit allocation.
Voice & video metrics
Section titled “Voice & video metrics”If your organization uses voice or video features, this section displays session-level statistics:
- Active voice sessions — number of currently active voice/video bot sessions.
- Total sessions — aggregate count over the selected time range.
- Average session duration — mean length of voice/video interactions.
- Session completion rate — percentage of sessions that ended normally (versus dropped or errored).
Filtering and time range options
Section titled “Filtering and time range options”All metrics on this page support flexible filtering:
- Time range selector — choose from preset ranges (last hour, 24 hours, 7 days, 30 days) or define a custom date range.
- Agent filter — narrow metrics to a specific agent or set of agents.
- Tool filter — focus on a particular tool or tool group.
- Status filter — show only successful, failed, or in-progress executions.
Filters apply across all sections on the page simultaneously, so you get a consistent view of the selected slice of data.
Exporting data
Section titled “Exporting data”You can export metrics data for external analysis or reporting:
- CSV export — download the current filtered view as a CSV file. Available for all table views (runs, tool calls, token usage).
- Time-range exports — export data for a specific date range, independent of the current page filter.
Next steps
Section titled “Next steps” Dashboard Overview Return to the main dashboard for a high-level view of system health.
Managing Agents Configure agents to optimize the metrics you see here.
Credits & Billing Understand how LLM token usage translates to credit consumption.