It writes. You edit.
Describe what you want to say in plain words, as you would to a colleague. Watch what happens next — this is not a video, it's the transcript of a real generation.
3,000 emails a month free · no credit card
A generator returns text. An assistant answers you.
If the result doesn't work, you don't start from scratch: you ask for a correction and it revises its own text. The previous exchange is kept — you never have to re-explain the context.
What happens between your sentence and the letter
It understands, even when poorly written
A dictated request arrives with typos, merged words, missing punctuation. That's no reason to refuse: it does its best with what it understood, and never sends a question back instead of a text.
- Dictation
- Typos tolerated
It checks outside
A news item, a recent figure, a day of the week: it searches the web rather than inventing, and announces what it found. That's the difference between a plausible letter and an accurate one.
- Web search
- Dates verified
It writes in blocks
What comes out is not a wall of code: it's the same blocks you would place by hand. The model decides the content, the system decides the form — that's what makes a letter arrive intact in Outlook.
- Six blocks
- No coding
It corrects you too
The request: 'Announce our masterclass on Thursday 31 December 2027 at 9 p.m.' 31 December 2027 does not fall on a Thursday.
What the assistant replied
Assistant
December 31, 2027 is a Friday, not a Thursday. Date corrected.
Replace [event title] with the name of your event.
The correction was not asked of the model: it is calculated. Every date written in full is checked against the calendar before the model sees anything, and the real day is imposed on it.
A language model gets the day of the week wrong with perfect confidence. So we don't ask it.
The voice belongs to the workspace, not the account
An agency writing for a law firm and a sports brand doesn't have one voice — it has two. Tone, register, emojis and length are set once per workspace, and each workspace keeps its own.
These are defaults, not restrictions. Emojis set to 'none' and you ask for 'add an emoji to the subject'? It adds one. A written instruction always overrides a saved setting.
The assistant is not capped on length, number of exchanges, or the right to search the web: what bounds it is a monthly budget included in your plan. Budget details by plan.
Frequently asked questions
Can I write without it?
Yes. The editor works on its own, with blocks to place or a large text area. The assistant is one way in, not the only way.
When the budget runs out?
The assistant stops, and nothing else. Your campaigns still send, the editor still works. The counter resets at the start of next month.
Where does the text I give it go?
To our model provider, for the duration of the generation. What is stored after that is the campaign — with us, in Europe.
Can it pick up a letter already written?
Yes. Open an existing campaign and ask for a change: it works on what's there rather than starting over.
What if it gets a fact wrong?
It happens — proofreading is not optional. It searches the web rather than inventing, and leaves a visible bracket rather than filling a gap. The pre-send check refuses to let a bracket that was never filled go out.
Try it on your real list
Three thousand emails a month free, no credit card. The assistant is included from the free plan.