Sit a full case with Dr. Ora, who questions you the way a real examiner does and tells you nothing until the end. Afterwards you get feedback on every answer you gave.
This describes what Mettle actually stores. It is a plain description of the software's behavior, not a legal document, and it should be reviewed by the department before the tool is offered outside it.
Cases you save, an exam left unfinished, and any API key you enter are stored in your browser on this device. They are not uploaded. Clearing your browser data removes them.
An anonymous transcript of every exam is saved so bad cases and bad examiner behavior can be found and fixed. Anonymous means the transcript carries no name, email, or account id: it records that an exam happened and how it went, not who sat it. This used to apply only to exams run on the department's key; it now covers exams run with your own key too. Every exam also records that it happened: the mode, the examiner tier, the specialty, and which cohort code was used. Ratings you give a case, and anything you send through the feedback box, are stored with what you type.
Running an exam sends the case text and the conversation to Anthropic, which provides the examiner and the grading. Spoken answers are sent to Groq or Omi to be transcribed, and the examiner's voice is generated by Deepgram. Voice Pro is different: it sends your microphone audio to OpenAI, which does the listening, the examining, and the speaking in one live session. Those companies process what they receive under their own terms.
If accounts are enabled, signing in stores your email address so your work can be attached to you. Saving a result to your account is a choice you make after each exam, not a default.
Plain terms of use, to be reviewed by the department before this is offered outside it.
Mettle is a practice tool for anesthesiology residents. It is not affiliated with, endorsed by, or associated with the American Board of Anesthesiology, and its cases are not actual examination material. Practicing here implies nothing about how you will perform in a real examination.
Cases and answer keys here are teaching material, and some are AI-generated and labelled as such. They are not clinical guidance and must not be used to make decisions about a real patient. Where an answer key is marked AI-generated it has not been verified by faculty.
Do not submit cases from commercial question banks, review courses, or other copyrighted material. Do not share your access code outside the group it was issued to. If you bring your own API key, its usage is billed to you.
This is provided as-is, without warranty. It is early software and it will sometimes be wrong or unavailable.
gsk_) and paste it above.Put the patient information first: age/sex, HPI, PMH, medications, vitals, exam, labs. Then your numbered questions (or “Phase 1 / Intra-operative” sections) below it. The app automatically splits the stem the candidate sees from the exam questions.
Already have the three-part ## PART 1: / ## PART 2: / ## PART 3: generator format? It’s split exactly, with no AI cost, and the confidential examiner script (PART 3) stays hidden from the stem.
Copy the case-generator prompt below, paste it into Claude or ChatGPT along with your high-yield topics, then paste the generated case back into the box above.
Before the exam begins you are given the case stem and 10 minutes to read it. Nothing is asked of you during this time. Use it to chart the patient in your own head and outline your approach: the problems, what you would want to know, and where you expect the trouble to be.
In Mettle this is the reading screen, with its own 10:00 clock. The exam clock does not start until you press Begin Exam.
The exam itself runs 35 minutes and comes in one of two shapes. You do not pick which. Choose Surprise me when you start a case and Mettle draws one at random, the way the real exam gives you what it gives you.
The stem is short. This patient has not been fully worked up, so part of what is being tested is what you would order and why. Ask for a study and the examiner gives you the result, then asks what you make of it.
The stem is long. It is the same patient, already worked up and optimised, and the exam opens heading into the OR. There is no credit here for deciding to order an echo: you were handed the echo, and you are answerable for every value in the stem from the first question.
The last 10 minutes of either format leave your case behind. Grab bag questions come from a separate pool and are on topics unrelated to the patient you have just managed. They are short, and they move quickly.
One question at a time, and no reaction to your answer. A real examiner does not tell you that you are right, does not tell you that you are wrong, and sounds exactly the same either way. Dr. Ora is built to do the same, so do not read anything into the tone. Silence is not disapproval.
If your reasoning takes the case somewhere else, the examiner follows you rather than correcting you, and steers back with a clinical turn of events rather than by telling you to go back. Time is finite, so a question you have already answered will not be asked again.
This page describes the format Mettle simulates and the timings it runs on. It is a study aid rather than an official ABA publication. Check the current examination content outline and candidate materials from the Board itself for anything that matters.
One account per resident. You need an email address and the signup code your program gave you. The code is what grants access, so treat it as yours rather than something to forward on.
You are signed in as soon as the account is made. After that, Sign in from the same place on any device, and your saved cases and results follow you.
Forgotten the password, or never set one? On the sign-in screen choose to have a code emailed to you, sign in with it, and then set a password under Settings so the next time is ordinary.
Signing in matters more than it looks. Your Voice Pro allowance is counted per account, and results you choose to keep are attached to your account rather than to the browser you happened to use.
New to the oral exam? Exam Format in the sidebar lays out what the 35 minutes actually look like, and the 10 minutes of reading time before them.
Pick a case in Case Library, then start it from Practice. You answer out loud or by typing, and Dr. Ora keeps to the pacing of a real exam.
Pause stops the clock and mutes your microphone. The break is written into your transcript rather than hidden, so the timings still mean something. If you close the tab mid-exam, you will be offered the chance to pick it up again for up to a week.
After a case you can ask for the answer key on its own, without being graded. You will see the answers for the questions your exam actually reached. Keys marked as AI-generated were drafted by Claude and are a study aid, not authority.
Voice Pro is a live spoken conversation and the closest thing to sitting the real exam. It is also much the most expensive to run, so everyone gets three included. After those you can add your own OpenAI key under Settings and carry on.
Voice sends your spoken answers turn by turn, and Text is the same examiner by typing. Both cost cents rather than dollars, so use them freely. Record captures a live mock oral with a real examiner and transcribes it afterwards, and Review grades a transcript you already have.
If Voice Pro will not start, it is almost always the microphone. Your browser asks once for permission. If you dismissed that prompt, click the icon in the address bar, allow it, and start the exam again.
Every exam is transcribed, and the transcript is kept so bad cases and bad examiner behavior can be found and fixed. Those transcripts carry no name, email, or account id. What is recorded against your name is only that an exam happened: the mode, the tier, the specialty, and which cohort code you signed up with.
Your scores and feedback are saved only if you choose Save to my account on the results screen. Nothing in the app shows those to anyone but you, and there is no program-director or cohort view. That is deliberate: you learn more from a session you were willing to fail.
Cases you save, an unfinished exam, and any API key you enter stay in this browser and are not uploaded. The full detail is on the Privacy page in the footer.
Use the Feedback button in the corner of any screen. It reaches us directly.
Mettle is an oral board simulator for anesthesiology residents. You practice answering out loud, on the clock, against an examiner that behaves the way a real one does. The AI examiner is called Dr. Ora.
Margaret Bowman-Chiu and Hamza Yunis, of the Department of Anesthesiology and Perioperative Medicine at the University of Louisville.
Mettle is funded by the Sessler Award from the University of Louisville Department of Anesthesiology and Perioperative Medicine. That funding is what keeps the examiner free for residents to use.
Knowing the material and saying it out loud under pressure are two different skills, and only one of them gets practiced regularly. Out-loud practice usually means finding a faculty member with a free hour, and there are never enough of those hours to go round. The idea here is to give you as much spoken practice as you want, whenever you want it.
An oral board examiner is a particular kind of difficult. They give you nothing: no hint that you are on the right track, no warmth when you flounder, no teaching until the case is over. That neutrality is most of what makes the real thing hard, and it is the first thing a conversational model gives up. Left to itself it starts helping, tells you when you are right, teaches mid-question, or drifts off the case.
So the examiner here is held to those rules deliberately, with a detailed brief for how it should behave and checks that test it against that brief rather than trusting it to stay in character. The point is not that the model is clever. It is that it stays cold, keeps to the clock, and makes you commit to an answer, which is the part you actually need to rehearse.
This is meant to be a free resource and we intend to keep it that way. What we are working towards is a library of vetted cases with humans in the loop, reviewed by people who know the material rather than generated and published unchecked.
We are still working out the bugs. We would rather you had the chance to practice now than wait for us to polish it, so if something breaks or could be better, use the Feedback button in the corner of any screen and tell us.