Custom Dashboards

AgentMark Custom Dashboards let you create configurable monitoring dashboards with drag-and-drop widgets. Track the metrics that matter most to your application with multiple visualization types, pre-built templates, and flexible filtering.

​

Templates

Start from a pre-built template or create a blank dashboard. Select a template from the gallery to instantly populate your dashboard with relevant widgets. AgentMark provides the following templates:

• Overview — high-level view of request volume, cost, errors, and latency with stat cards and time series charts
• Cost Analysis — detailed cost tracking with total cost, average cost per request, cost over time, top models by cost, and token trends
• Performance — latency percentiles (P50, P95, P99), error count, error rate, and average latency

You can customize any template after creating the dashboard.

​

Widget types

Each widget displays a single metric using one of the following visualization types:

• Stat card — a single metric value, optionally showing the change over the previous period (direction: up, down, or flat)
• Line chart — a metric plotted over time, useful for trends and patterns
• Bar chart — a metric grouped by a dimension such as model or user, useful for comparisons and top-N rankings
• Area chart — similar to a line chart with a filled area beneath, useful for volume-based metrics

​

Available metrics

AgentMark Custom Dashboards support the following built-in metrics: Volume and usage:

• request_count — total number of requests
• unique_users — count of distinct users
• total_tokens — total tokens consumed
• avg_tokens — average tokens per request

Cost:

• total_cost — total dollar cost
• avg_cost — average cost per request

Errors:

• error_count — total number of errors
• error_rate — percentage of requests that failed

Latency:

• avg_latency — average response time in milliseconds
• p50_latency — 50th percentile latency
• p95_latency — 95th percentile latency
• p99_latency — 99th percentile latency

Rankings:

• top_models — model usage ranked by request count

Derived metrics are also available. These are computed as ratios of built-in metrics, such as cost per request, tokens per request, cost per user, and success rate.

​

Creating a dashboard and adding widgets

1. Navigate to your app and click the Dashboards tab
2. Click + New Dashboard or select Use Template from the template gallery
3. The dashboard opens with your widgets — click + Add Widget to add more
4. Configure the widget: choose a title, metric, visualization type, and optional group-by dimension
5. Click Add Widget to place it on the dashboard

​

Customizing widgets

Each widget can be configured with the following options:

• Metric — the data point to display (see Available metrics)
• Visualization — how to render the data (stat, line, bar, or area)
• Group by — an optional dimension to segment the data by model, user, time period, or a custom metadata key
• Time granularity — hour, day, or auto (automatically selected based on the date range)
• Filters — narrow the widget’s data by model, user ID, status, or metadata values

Drag widgets to rearrange the layout. The grid supports up to 12 columns, and widgets can span any width and height.

​

Dashboard settings

• Default dashboard — mark a dashboard as the default to load it automatically when you visit the Dashboards tab
• Global time range — set a time range that applies to all widgets in the dashboard (24 hours, 7 days, 30 days, 90 days, or custom)
• Limits — each app supports up to 10 dashboards, with a maximum of 20 widgets per dashboard

​

Next steps