For therapists and counsellors

For therapists and counsellors: the enquiry reply that has to sound like a person

The clinical work is in the room. The email is the admin around it, and in private practice there is nobody else to do it: the enquiry that asks about fees, availability and how you work, the waitlist update, the reschedule, the cancellation policy explained again with the exception you have decided to make, the monthly statement with the codes an insurer wants. It is the least clinical part of the job and it lands in the gaps between sessions.

The enquiry reply is the one that matters most and gets the least time. Someone who has finally written to a therapist can tell the difference between a person answering and a clinic answering, and a reply that reads like a brochure is often the last one they open.

MailTwin is built the other way round. You type the points the reply needs to make, and it writes the finished email in your own words, copied from how you already write. You read it and send it.

What that looks like between sessions

Select the message in Apple Mail, type your notes, and MailTwin writes the complete email into the reply window — greeting, every point in full sentences; Mail appends your own signature.

You type	You get
Tuesday 5pm free from next month, 50-minute sessions, fee 90, first session is an intake, no obligation after it	An enquiry reply that answers the practical questions warmly and makes the first step feel small
still on the waitlist, nothing has opened yet, will write the moment it does, two colleagues here if they would rather not wait	A waitlist update that does not leave them hanging and gives a real alternative
missed Thursday, 24-hour policy would apply, no charge this time given the week they have had, same time next week	The cancellation email that states the policy, makes the exception you chose, and moves on
March statement attached, four sessions, includes the codes their insurer asked for, tell me if the format is wrong	A short cover note for the statement, in the register you use for admin rather than the one you use in session

None of those is clinical. That is deliberate. MailTwin is for the administrative mail around the work, and anything that needs your clinical judgement stays with you, because you read every draft before it goes.

Client correspondence, and where it goes

Whether an AI tool should touch client correspondence at all is a decision you make under the code you practise under, and MailTwin does not make it for you. What it can do is make the decision a clean one.

MailTwin has no server. There is no MailTwin account, and there is nowhere for your mail to be sent even if we wanted it. You choose the AI engine from nine. Three of them, Apple Intelligence on macOS 26 or later, Ollama, and a local OpenAI-compatible server (oMLX, LM Studio, llama.cpp), run on the Mac itself, so with those the message content stays on the Mac and is not transmitted. With a cloud provider, the request goes from your Mac straight to that provider under a key you supply and can revoke, together with a short sample of your own sent mail so the reply sounds like you. Voice learning reads sent mail locally and keeps its samples locally, under your own user account.

MailTwin does not claim compliance with any health-privacy regime, and no software can make that claim for a practice on its own. Your professional body's code, the rules you work under, and whatever consent your clients have given decide whether a cloud provider is acceptable for their correspondence. If the answer is no, the on-device engines are a complete answer rather than a reduced version of the app.

The voice is the point

MailTwin reads your own sent mail, on your own Mac, and copies your habits: how you open, how warm you are, how much you say and how much you leave unsaid. It never writes a sign-off of its own; Mail appends your signature. It learns separately for every mail account, so the practice address and the personal one keep their own register.

The draft is a first pass by something that has learned your habits. It is not a decision made on your behalf, and in this work that distinction matters more than in most. Read it before it goes.

Fees, times and the cancellation policy

When MailTwin rewrites a draft you have already written, a validator checks the result before it reaches you and rejects it if a number, a date, an amount, a link or an email address has changed. That runs every time. What it cannot do is judge whether a new sentence is true; type 5pm when you meant 6pm and you will get 5pm in a well-written sentence. Nothing sends without you.

The honest version of the time saving

Christopher Naas, who makes MailTwin, says it saves him an hour or more on a busy email day. That is his experience, not a promise about your practice, and anyone quoting a precise figure is guessing. What holds is the shape of it: the time was never in the one email that needed thought. It was in the fifteen routine ones written between sessions.

What it will not do

It does not read the inbox on its own, does not answer anyone without you, and never sends; there is no autonomous mode. It does not connect to a practice management system, a booking tool or your notes, and it does not know who is on the waitlist unless you type it. It is macOS 13 or later only, Apple Mail only, with no iPhone or iPad app. It needs two macOS permissions on the first run, Automation for Mail and Accessibility, which is a few minutes in System Settings and cannot be automated away, because Apple deliberately does not let an app grant itself access to your mail. And if your replies are mostly one line, it will not help much. It is for the paragraph-long enquiry answer, not the thumbs-up.

Common questions

Is it appropriate to use an AI tool on client correspondence at all?

That is your call, made under the code you practise under, and MailTwin does not make it for you. What it offers is a way to keep the content on the Mac while you make it: three of the nine engines, Apple Intelligence on macOS 26 or later, Ollama, and a local OpenAI-compatible server, run on the Mac itself, so with those the message is not transmitted. With a cloud provider the request goes to that provider under your own key. MailTwin does not claim compliance with any health-privacy regime; your own rules, and any consent your clients have given, apply.

Will it sound like me or like a clinic?

Like you. It reads your own sent mail, on your own Mac, and copies your habits: how you open, how warm you are, how much you say and how much you leave. It never writes a sign-off of its own; Mail appends your signature. You read every draft before it goes, and for this work that matters more than in most.

Could it change a fee or a session time?

When it rewrites a draft you have already written, a validator rejects the result if a number, date, amount, link or email address has changed. What it cannot do is judge whether a new sentence is true, so a wrong time you typed comes back written out neatly. Read every draft before it goes. Nothing sends on its own.

Does it read my inbox or send anything itself?

No. It does not read the inbox on its own, it does not answer anyone without you, and it never sends. Every message is a draft in your reply window that you read, change and send yourself. There is no autonomous mode and no scheduled sending.

Is it a subscription?

No. It is a one-time licence, with a 30-day full trial and no card required to start. The three engines that run on the Mac cost nothing to run. If you pick a cloud engine you pay that provider for what you use, typically a few cents a day at ordinary email volumes, and the app shows a running total.

Try it on a month of enquiries

Thirty days, every feature, no card. A month of real enquiries tells you more than any description.

Download MailTwin →

Other guides

Weighing it against something else? Best AI email assistant for Mac in 2026 compares six options and says who should pick each one.