Odile Labs
Call (561) 375-2995Book a demo
(561) 375-2995Call

Odile Agent Ops · for agencies and teams running AI voice agents

Keep your voice-agent tools. We test them every week.

Odile calls each AI voice agent you run, on any vendor, with 5 scored scenarios every week: a booking, an unavailable time, a cancellation, a language switch and a request that should go to a person. You get the transcripts, a pass or fail report and an email alert within one business day when a scenario fails.

QA $99 a month · Partner $499 a month · Scale $1,499 a month · by card, cancel any time

Plans

QA

$99 a month

  • Up to 3 agent lines
  • 5 scored scenarios per line, every week
  • Transcripts and a pass or fail report
  • An email alert within one business day of a failed scenario
  • A console for all your lines, and a private report page per line to share with a client
  • Retest a failed scenario after a fix (it counts toward that week’s calls)

Start QA · $99 a month

Partner

$499 a month

  • Up to 10 client lines
  • Everything in QA
  • One report per client each month
  • A named escalation channel
  • One standard setup a month of an Odile line, Odile Direct or Odile Gate where it fits

Start Partner · $499 a month

Scale

$1,499 a month

  • Up to 50 client lines
  • Everything in Partner
  • One failure feed across the fleet
  • Up to 3 standard setups a month

Start Scale · $1,499 a month

Running more than 50 lines? Start with the team pilot (30 days, $2,500).

A real report, from our own test line

Five real calls in the week of 2 November to Odile Test Home Services, a fictional receptionist we run on ElevenLabs to test ourselves. No client data.

  • A booking: pass.
  • An unavailable time: fail. The caller gave 561-555-0134. The receptionist read it back three times as “five five one five five five zero one three four”, so the office would have called the wrong number.
  • A cancellation: pass.
  • A language switch: pass.
  • A request that should go to a person: pass.

Our scorer first passed all five. A second read of the transcripts caught the wrong digit, and we changed the scorer so a wrong callback read-back fails the call. These results are the five calls replayed through the changed scorer, four times each, with the same result every time. A customer report also carries the transcripts.

What you give us

  • The numbers of the agent lines: landline or VoIP numbers you own or are authorized to test.
  • Test windows, and the outcome each scenario should reach.
  • Where failure alerts should go.

What it is, and what it is not

  • Scheduled testing: calls on a weekly schedule, scored automatically, and a failed scenario is checked before the alert goes out. It is not continuous monitoring of live calls.
  • No uptime or response-time guarantee, no certification or compliance claim, and no promise that every failure is caught.
  • Your voice vendor’s call charges, and any Odile usage at published prices, are extra.

How Agent Ops compares with Coval, Hamming and Loops: who runs the tests and what each costs.

Clients on LeadConnector?

Any client that needs a receptionist can install After Hours by Odile from the LeadConnector App Marketplace: see how. Developers can run one scored test of their own agent for $1 with x402: the test-call API.