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.
What did you just agree to in that meeting?
Paste a meeting transcript and this finds every promise made in the room, yours and theirs. Each one comes back with who owns it, the deadline if the sentence gave one, and a band for how likely it is to slip. Speaker names are read from the lines, not guessed. It is a free sample from Mailient.
- Free
- No account
- Your transcript is not stored
- Answers in seconds
What you paste
Three steps, no sign-up
Paste the transcript
Any export with the speaker's name at the start of each line: Zoom, Google Meet, Otter, Granola, Fathom, Fireflies, or notes you typed yourself. Timestamps are fine.
Say which speaker is you
That is what splits what you owe from what you are owed. The names are read from the lines in your browser, so nothing has been sent yet.
Read what was promised
Each promise comes back with a risk band, the deadline if the sentence had one, whether it is clear who owns it, and the exact words behind each call, underlined.
What you get
What each part of a result is
Every field on a result is either read from the transcript or judged by a model. The table says which, so you know what to check.
| Part | Where it comes from | What it tells you |
|---|---|---|
| Who said it | Read from your text | The name on the line. Attribution is parsing, so it cannot tell you that you owe something you never said. |
| Is it a promise | Model judgement | Whether the speaker took on an obligation. Questions, suggestions nobody accepted and “someone should…” are dropped. |
| How firm | Model judgement | Firm, hedged or vague. A vague line, or one said as “we”, shows the owner as unclear instead of naming the speaker. |
| Deadline | Read from your text | A day, date or phrase in the promise's own sentence. If that sentence names none, the card says Not stated. |
| Hedges | Read from your text | Softeners in the sentence, such as “probably” or “hopefully”, underlined in amber. |
| Admissions | Read from your text | Phrases like “I keep forgetting” or “I'm behind”, underlined in red. An acknowledged debt is usually the most fragile promise in the room. |
| Risk band | Model judgement | Likely to slip, could slip or on track. A judgement about how the promise was made, not a forecast of what happens next. |
Reading it
How to read the result
Start with your own promises. Say which speaker is you and the list splits into what you owe and what you are owed. Each card shows the line it came from, so you can check any call against what was actually said.
The risk band describes how a promise was made, not what will happen. Likely to slip usually means no date, a hedge such as “probably”, or an admission such as “I keep forgetting”, and those words are underlined. On track means a named date, a clear owner and no hedging.
“Not stated” in the deadline field means the promise's own sentence named no day or date. The tool will not borrow one from a nearby line, so add the date yourself if you agreed one.
Background
Why it works this way
What counts as a commitment in a meeting?
A commitment is a line where a speaker takes something on: “I'll send the draft Thursday”, “I'll ping legal today”. A question, a suggestion nobody accepted, or “someone should look at that sometime” is not one, because nobody owns it.
The ones that matter most are usually the untidy ones: an acknowledged debt (“I owe him an intro, I keep forgetting”) or a promise with no date at all. Those are what this tool flags first.
Why read a transcript at all?
The transcript that would tell you what you agreed to already exists, and nobody reads it, because reading a 4,000-word transcript costs more than the meeting did. This reads all of it in a few seconds and hands you back the list.
Limits
What it cannot tell you
Every tool here is a first pass with edges. These are this one's.
- It cannot tell you whether a promise was kept. The band describes how the promise was made. Nothing here follows up afterwards.
- Deadlines come from one sentence only. If someone says “Thursday” in the line before, the promise that follows still shows Not stated. A wrong deadline does more harm than an empty one.
- It is only as good as the transcript. Speech-to-text mistakes carry through, and a transcript with no speaker names cannot be read at all. The page says so as soon as you paste, rather than guessing.
- Long meetings are cut. One run reads up to about 400 lines, roughly a two-hour meeting.
- Runs are limited. Twelve an hour from one connection, so a free tool cannot be run thousands of times.
Beyond the sample
When this is not enough
A transcript shows what was promised in one meeting. It cannot tell you whether the follow-up email ever went out, or whether the other side replied.
Mailient does not read meeting transcripts. On your Gmail, the nearest thing is its Today view, which lists the threads you sent 3 to 14 days ago that got no reply, alongside the emails that need a decision today and today's meetings. It drafts a follow-up in your writing voice when you ask, and by default sends nothing until you approve it.
Your data
What we keep
Read off the code that does it, not written from memory.
- In your browser
- Speaker names and line breaks are worked out as you paste.
- Sent to our server
- The transcript, when you press the button. It is held in memory for that one request, up to about 400 lines.
- Sent on to a model
- Yes. Each line is put to three questions by Jev, a decision model, through OpenRouter.
- What we keep
- Nothing you paste. A request counter tied to your IP address (12 runs an hour) holds no content.
Read from app/api/tools/commitments/route.ts and lib/rate-limit.ts.
Questions
Questions people ask, answered plainly
6 answers, grouped. Each one is written against what the tool actually does.
Using it
Which transcripts work?
Any export where each line starts with the speaker's name: Zoom, Google Meet, Otter, Granola, Fathom, Fireflies, or notes you typed yourself in the form “Sarah: I'll send it Thursday”. Timestamps are fine and are ignored. A block of text with no names cannot work, because everything depends on knowing who said what, and the page tells you that the moment you paste.
How it decides
How does it know what counts as a promise?
Each line is put to a decision model as three plain questions: is the speaker taking on an obligation, how firm is it, and how likely is it to be quietly forgotten. Questions, opinions and suggestions nobody accepted are not obligations, so they are dropped. An acknowledged debt, like “I owe him an intro, I keep forgetting”, counts, and is usually the most fragile one in the room.
What does “likely to slip” actually mean?
It is a judgement about how the promise was made, not a prediction of what happens next. A commitment with a named date, a clear owner and no hedging reads as solid. One with no date, or one where the speaker admits they have already forgotten it once, reads as fragile. It is shown as a band rather than a percentage because there is no outcome data behind it, and a number like 68% would imply a precision it has not earned.
Accuracy
How accurate is it?
We have not scored it against a labelled set of real meetings, so we will not quote an accuracy percentage. What we have tested is repeatability. In September 2026 we ran one 59-line meeting through it five times and got the same 15 promises every time, with risk scores that moved by at most 0.06 on a scale of 0 to 1. Treat it as a fast first pass to check against what you remember, not as a record of the meeting.
Your data
Do you store my transcript?
No. Speaker names and line breaks are worked out in your browser. When you press the button the lines go to our server, which holds them in memory for that one request and sends them on to the model to answer the three questions per line. Nothing is written to our database. The one thing we do keep is a request counter tied to your IP address, which limits runs to 12 an hour and holds no content.
Mailient
Is this what Mailient does?
Not exactly. This page finds the promises in a transcript you paste, and Mailient does not read transcripts. It works on your Gmail: its Today view shows the emails that need a decision today, today's meetings, and threads you sent that got no reply; it drafts a reply in your writing voice when you ask; and by default it sends nothing until you approve it.
More free tools
Try another one
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.
This is the sample. Mailient does this on Gmail.
This tool runs on text you paste. 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.