Read the call log

Every call GodSetter places or answers is written to the Calls page: who was called, how it ended, what was said, what it cost and which tools the agent used while talking. This is where you check a specific conversation — the Dashboard tells you how the week is going, the call log tells you what happened on one call.

All times on both the list and the detail are shown in the Europe/Rome time zone, as the header line under the page title states.

The list

Rows are ordered by start time, most recent first, 25 per page. Each row carries:

The Open call detail button at the end of the row goes to the same place as the lead cell. Anything GodSetter does not have is rendered as a dash — no value on this page is estimated.

Filters and pagination

Five pills filter the list on the call's outcome: All, Booked, Callback, No answer, Failed. The number on each pill is a real count over your whole call log, not just the page you are looking at, so the pills double as a quick tally.

Under the table, 1–25 of 120 tells you where you are, with Previous page and Next page to move. The filter and the page are both in the URL, so a filtered view can be bookmarked or shared with a colleague.

Inside a call

Opening a call gives you the full record. The header repeats the start date and time, the duration, the phone number (with Copy number), the outcome badge, the direction and — when applicable — the Test badge.

The Transcript panel lists the conversation turn by turn, each line marked AI or Lead. When Retell sent no transcript you get No transcript available for this call. instead of an empty box.

The right-hand rail holds three cards:

• Result — Outcome, Duration, Recording (available or a dash), Cost, Latency, and an Error row that only appears when the call recorded an error message.
• Tools invoked — see below.
• Metadata — Call ID, Direction, Status, Voice / LLM, From number, To number, Duration. The Call ID is Retell's own identifier for the call: it is what to quote in a support request.

Listening to the recording

When the recording has been archived, a player appears above the transcript. Press play, and click anywhere on the waveform to jump to that point (Go to a point in the recording). The elapsed and total times sit on either side, and the button on the right cycles the playback speed through 1.0×, 1.25×, 1.50× and 2.0× — the fast way to review a long call.

No player means no archived recording for that call. If recordings are missing in general, see Recordings are not archived.

Cost and latency

Both figures are measured, not estimated: they come from the cost and latency data Retell attaches to the call when it ends and when its analysis lands. A call that never received them shows a dash.

Cost is the total in USD, with a per-product breakdown listed underneath — one line per item Retell billed (voice infrastructure, the LLM, transcription, and any add-on you enabled), each with its own amount. That is where a call that cost more than you expected explains itself.

Latency (p50 / p90) is the end-to-end response time in milliseconds — the median and the slow tail of the same call. When component figures are available, a second line breaks it down: LLM … ms · Voice … ms · Transcription … ms. p50 is what the conversation felt like most of the time; p90 is the worst of it.

Tools invoked during the call

The Tools invoked card is the timeline of everything the agent actually did while talking — one line per invocation, in order:

→ get_free_slots (420 ms) → 6 slots

The name is the tool, the milliseconds are the measured duration, and the last part is the outcome: 6 slots, confirmed, slot taken, alternatives offered, 2 appointments, appointment not found, cancelled, rescheduled, tagged not interested, callback scheduled, ok, or error: … on a failure — failed lines are shown in red. A test chip marks an invocation that was simulated rather than sent to your CRM. When the agent used nothing, the card reads No tool invoked.

This is the fastest way to tell "the agent never tried to book" from "the agent tried and the slot was taken".

Test calls

Calls started from the agent tester, and calls to a test lead, carry a Test badge in the list and on the detail. They are real calls with a real cost, but they are kept out of your Dashboard KPIs and out of the per-version metrics, so testing never skews the numbers you report.

Next

• Manage your leads
• How outbound dispatch works
• Prompt versions, A/B metrics and rollback