MeetingsA few seconds
Meeting commitment finderWhat did you just agree to in that meeting?
Paste a meeting transcript. Get every promise made in the room, yours and theirs, with the ones most likely to be forgotten flagged first.
Free tools for people who live in email
These five tools are free and need no account: paste some text or type a domain, and the answer comes back in seconds, measured from your own text and public records. Each one is a small, working sample of the kind of reading Mailient does on Gmail.
- No account
- No card
- Answers in seconds
InboxInstant, then a second look
Inbox triage testWhich emails in this list actually need you?
Paste the subject lines in your inbox. Get them sorted into what needs you, what can wait and what a machine sent, with the reason for each.
SecurityAbout two seconds
Email spoofing checkerCan someone send email as this domain?
Type your domain. See whether a stranger can email your customers from your address, what that email would look like, and the exact records that stop it.
WritingInstant, then a few seconds
Writing voice analyserWhat does your writing voice actually look like?
Paste two emails you have sent. Get your writing fingerprint, measured from the text rather than guessed, and one new email in your voice.
WritingTwelve seconds a round
Spot the defaultWhich reply was written with a voice profile?
Two replies to the same email, both written by AI. One had a voice profile and one had none. You have twelve seconds to say which.
Being built next
Thread decision extractor
Paste a long email thread. Get what was actually agreed and who owes what.
It ships only if it does something a chat window is genuinely bad at.
How they are built
Measured first, judged second
Three rules every tool here has to keep. They are why the answers show their evidence.
Counted before it is judged
Whatever can be measured is measured in code: a sentence length, a DNS record, a deadline in a sentence. A language model cannot count and will invent a plausible number, so models are used only for judgement, and what they say sits beside the text that supports it.
Evidence you can check
Every call points back at its source: the underlined words in a transcript, the raw DNS record, the line from your own email. If a tool cannot show you why, it says less instead of more.
Honest about the limits
Risk is shown as a band because there is no outcome data behind a percentage. A missing DKIM key is reported as not found, not as “you have none”. The game says every email in it was written by AI. Each page lists what its tool cannot tell you.
Your data
Where what you give a tool actually goes
Read off the code that does it, one row per tool. Nothing is left out of the table because it is awkward.
| Tool | In your browser | Sent to our server | Sent on to a model | What we keep |
|---|---|---|---|---|
| Meeting commitment finder | Speaker names and line breaks are worked out as you paste. | The transcript, when you press the button. It is held in memory for that one request, up to about 400 lines. | Yes. Each line is put to three questions by Jev, a decision model, through OpenRouter. | Nothing you paste. A request counter tied to your IP address (12 runs an hour) holds no content. |
| Inbox triage test | The first sort, by pattern. It is on screen before anything is sent. | Your subject lines (up to 40, each cut to 200 characters), so the sort can be re-checked and the unclear lines picked out. | Only the lines the patterns could not place, and only if there are any. A small open model (Gemma) through OpenRouter. | Nothing you paste. A request counter tied to your IP address (10 model calls an hour) holds no content. |
| Email spoofing checker | Checks that what you typed looks like a domain. | The domain name. We read its public DNS records from our server. | None. The verdict is rules applied to the DNS records. | The domain, its grade and when it was checked, only if it receives mail; that is what gives the result its page. If you choose to be told when it changes, your email address as well, until you unsubscribe. |
| Writing voice analyser | The measurements: sentence lengths, hedging, greetings and sign-offs. They appear instantly, with no network call. | The emails, so a model can write the reading. Each is cut to 3,000 characters. | Yes, for the written reading only: a small open model (Gemma) through OpenRouter. | Nothing you paste. A request counter tied to your IP address (10 readings an hour) holds no content. |
| Spot the default | The whole game. Your best score is saved in this browser and nowhere else. | Only a request for a fresh round, carrying the topics of the last few rounds. Nothing about you. | Yes, to write new rounds while you play (Gemma through OpenRouter). You type nothing, so it sees nothing of yours. | Nothing about you. A request counter tied to your IP address (30 rounds an hour) holds no content. |
Questions
Questions people ask, answered plainly
11 answers, grouped. Each one is written against what the tool actually does.
About these tools
Are these really free, and what is the catch?
They are free, they need no account, and no card is asked for. The catch, stated plainly, is that they exist so you find out Mailient exists. Each one ends with an offer to try the product, and that is the whole business model of this page. It seemed better to say so than to pretend otherwise.
What decides whether a tool gets built?
One rule: it has to do something a chat window is genuinely bad at. An earlier version of this page had three reply generators on it, which was a mistake, because pasting context into ChatGPT does that job faster. Measuring, checking DNS, reading a transcript and testing your own judgement all qualify. Drafting does not.
What is coming next?
A thread decision extractor: paste a long email thread and get what was actually agreed and who owes what. It ships only if it clears the same bar as the rest.
What happens to what you paste
What happens to what I paste?
It depends on the tool, and the table above lists exactly where each one's input goes: what stays in your browser, what reaches our server, what goes on to a model, and what is kept. The short version is that the text you paste into the meeting, inbox and writing tools is never written to our database.
Do you store anything at all?
Two things. A request counter per tool, tied to your IP address, so nobody can run a free tool thousands of times; it holds no content. And for the spoofing checker, the domain you checked with its grade and the time, because every check has its own public page. If you choose to watch a domain, we also keep your email address until you unsubscribe.
Which AI companies see my text?
Where a model is used, the text goes through OpenRouter, a routing service, to the model that runs it: Google's Gemma for the writing analyser, the Spot the default rounds and the second look in inbox triage, and TypeSafe's Jev for the meeting commitment finder. The spoofing checker uses no model. Those providers have their own data terms, which we do not control.
Do you use what I paste to train anything?
We do not keep it, so there is nothing for us to train on. We cannot speak for the model providers' own policies, which is why the table says exactly which tools send text to a model.
Should I paste something sensitive?
Use your judgement, and edit out anything you would not send to a third-party service first. That is good practice with any tool that sends text to a model, ours included.
How they work
Why do the tools show numbers and evidence instead of an AI opinion?
Because a language model cannot count and will happily give you a plausible number that is not yours. Wherever something can be read or measured directly, such as sentence length, a DNS record or a deadline in a sentence, the tools do that in code. A model is used only for judgement, and its answer is shown next to the text that supports it, so you can check it.
How accurate are they?
None of them has been scored against a labelled test set, so we do not quote accuracy percentages. Each page says what its tool cannot tell you and shows the evidence for each answer. Where we have measured something, such as the commitment finder returning the same promises across repeated runs, the page says so and gives the date.
Mailient
Is this what Mailient does?
It is a small piece of it. Each tool does one job on something you paste or type. Mailient does the same kind of reading on your Gmail: it shows what needs a decision today, drafts in your voice when you ask, and by default sends nothing until you approve it.
This is the sample. Mailient does this on Gmail.
Each tool runs on something you paste or type. Mailient does the same kind of reading on your Gmail: it shows what needs a decision today, drafts in your voice when you ask, and by default sends nothing until you approve it.