Monitoring
Monitoring shows what your agents are doing right now and how healthy they've been over a date range. Open it from your project under Observe → Monitoring.
Use Monitoring to check on live traffic and spot problems. For trends over time, cost and session end reasons, use Analytics. For what callers ask, use Insights.
Pick what to look at
The toolbar at the top filters everything on the page except the live band:
- Date range: choose 24h, 7d or 30d, or set your own start and end dates. The default is the last 7 days.
- Agent: show one agent, or All agents.
The Channel filter and Compare to previous aren't available yet.
Right now
The dark band at the top lists the calls that are live at this moment, across every agent in the project.
| Figure | What it means |
|---|---|
| Live now | Calls in progress. |
| Longest call | How long the longest live call has been running. |
| Agents on calls | How many different agents are on a call. |
| Channels | How many channels (phone, widget and so on) are in use. |
Each live call shows the agent, the channel, when it started and how long it has run. Click a call to open its live transcript in Calls & transcripts.
The band refreshes every 30 seconds while Auto-refresh is on. Turn it off to freeze the list, and click the refresh button to update it by hand.
Key metrics
The row of tiles covers the selected date range and agent.
| Tile | What it shows |
|---|---|
| Live now | Active sessions, the same count as the live band. |
| Sessions today | Sessions started today, with the total for the range below it. |
| Tool success rate | Share of tool calls that succeeded, and how many tool calls were made. |
| Avg duration | Average session length. |
| Latency p50 / p95 | Agent response time. p50 is the typical response; p95 is the slowest 1 in 20. |
| Errors | Errors logged in the range. The tile turns red when there are any. |
Latency and errors
- Response latency charts the p50 and p95 response times for the range. If p95 is much higher than p50, some turns are slow: check tool calls and knowledge base lookups in Logs.
- Errors by type groups the errors in the range, with the count and share of each type. Click View in logs to see the individual errors.
Sessions per day
A bar chart of sessions for each day in the range (up to the last 30 days of it). Use it to spot a sudden drop, which often means a channel or number stopped reaching your agent.
Agent performance
A table with one row per agent: sessions, successful and failed tool calls, average duration, and how many calls the agent has live now. Click an agent's name to open it.
Coming soon
The page shows placeholders for Call lanes (one bar per call, coloured by how it ended) and Uptime per channel. These aren't available yet.
Common tasks
- A caller says the agent is slow: check Latency p50 / p95, then filter Logs by that agent and look at tool call durations.
- Errors went up: open Errors by type, click View in logs, and filter by Error.
- Listen in on a live call: click it under Right now to follow its transcript.
- No data: the page shows "No activity in this period". Pick a wider date range or All agents.
Plans and access
Live calls and the metrics are separate plan features. If your plan doesn't include one, that part of the page shows an upgrade notice instead. See Usage and billing.