Platform
Dashboards
Combine metrics, charts, filters, and activity feeds to monitor agent behavior over time.
Last updated
On this page
Custom Dashboards
The Observability tab in your project provides customizable dashboards for monitoring agent performance over time. Create multiple dashboards and arrange their widgets in a drag-and-drop grid so each view answers one operational question.
Widget Types
Stat Cards
Show one metric with an optional breakdown: total runs, success rate, failures, tool calls, tokens, or token cost.
Area Charts
Plot agent or connector runs, token usage, token cost, or model usage as toggleable time series.
Bar Charts
Rank agents by runs, tokens, token cost, or judge score; context values by runs, tokens, or cost; models by tokens or cost; connectors by executions; and judges by score.
Logs Lists
Add a recent activity feed for agent or connector runs, with an errors-only option and links to run details.
Input Widgets
Bind dashboard-level text or select values that other widgets can share. Select options can be static or loaded from project database collections.
Dashboard Features
- Multiple dashboards: separate cost monitoring, reliability, and debugging views.
- Date range picker: change the time window for every widget with a preset or custom range.
- Auto-refresh: refresh live ranges every 10, 30, or 60 seconds according to the selected range; fixed past ranges do not refresh.
- Per-widget filters: scope a widget to particular agents or connectors.
- Shared input variables: define one input and reuse its value across charts, stats, and logs.
- Drag-and-drop layout: arrange and resize widgets in edit mode.
- AI editing: add or revise widgets from a plain-language dashboard prompt.
- Default dashboard: use the preconfigured dashboard included in each project.
Agent-Level Observability
Open an agent from the Agents tab for a focused operational view. It combines that agent's configuration and recent executions without requiring a custom dashboard.
- Statistics: total runs, success rate, total runtime, token cost, and tokens, with average, p50, and p95 per-run views.
- Configuration: agent type, model, concurrency, and linked tools or agents.
- Runs table: an agent-scoped table with status, context, guardrail, deployment, and date filters.
- Manual trigger: send a payload to the agent directly from the dashboard.
Dashboard metrics and billing details
Usage and Billing documents token categories, cost calculation, plan limits, and billing periods.