Gerador de Email

What the Email Generator does

You tell it what you're trying to say. It writes the email. That's the whole interface.

Pick an intent — replying to a request, declining politely, following up on a thread, cold-pitching a stranger, asking for an introduction, breaking bad news. Set a tone (formal, casual, direct, warm) and a length (one-line reply, short paragraph, full message with context). Hit Generate. The draft appears, ready to copy into Gmail, Outlook, or whatever client you live in.

It's not a replacement for thinking. It's a replacement for the twenty minutes you'd otherwise spend rewriting the same email three times because you couldn't get the opener right. The model handles the boilerplate — the greeting, the structure, the closing — and you spend your attention on the one or two sentences that actually carry the message.

How to use it

1. Describe the email you want to send. One sentence is enough: "decline a meeting next Tuesday but offer to do it the following week" or "follow up on the proposal I sent two weeks ago."
2. Pick the tone. Formal for clients and external stakeholders, casual for coworkers you know, direct for senders who don't want pleasantries, warm for relationship-building.
3. Pick the length. A one-line reply is sometimes all a thread needs. A detailed paragraph is for cold outreach or complex updates.
4. Add any specific details: names, dates, the prior context the email is responding to.
5. Generate. Edit. Copy into your email client. Send.

The whole loop takes under a minute for most emails. Compare that to opening a blank compose window and staring at it until something passable comes out — which, depending on the email and the day, can take half an hour.

Worked example: declining a meeting, offering an alternative

Here's a typical input. Someone has invited you to a 60-minute meeting Tuesday at 2pm. You can't make it, but you genuinely want the conversation to happen — you just need to push it. The intent box reads:

Tone: warm. Length: short paragraph. Generated output:

That email took the model about two seconds to write. It would take a careful person five to ten minutes to land the same balance — apologetic without grovelling, proactive about rescheduling, warm enough to protect the relationship. The AI doesn't do anything magical; it just doesn't agonize.

Style options and when to use each

The four tone settings produce noticeably different output for the same input. They're not cosmetic — picking the wrong tone is one of the most common reasons a draft feels off.

If the first draft feels wrong, the fastest fix is usually switching tone and regenerating rather than rewriting the draft you have.

Length presets and how to pick

The same intent can produce a 12-word email or a 180-word email depending on length setting. Most emails should be shorter than they are.

• One-line reply. "Confirmed — see you Tuesday." "Got it, will follow up by EOD." "No, thank you — not the right fit right now." For threads where the social contract demands a response but there's nothing substantive to add.
• Short paragraph (40–80 words). The default for most workplace email. A greeting, the message, a close. The decline-the-meeting example above is in this range.
• Detailed (100–200 words). Cold outreach, status updates with multiple data points, proposals, anything where you're introducing context the recipient doesn't have. Past 200 words you're better off in a doc with a link.

A rule that holds up well: if you're sending the email to someone whose response will be "okay, thanks," you wrote too much. If you're sending it to someone whose response will be "wait, what?" you wrote too little.

Why this is faster than opening ChatGPT yourself

You can absolutely do this in ChatGPT. Open a new chat, type "Write me an email to my manager declining the all-hands meeting Tuesday because I have a doctor's appointment, but I want to read the recap," paste in any context, get the output. Works fine.

What this widget removes is the prompt engineering tax. The widget already knows what an email looks like. It already knows about openers, closes, subject lines, and tone calibration. You don't have to teach it any of that. You just describe the intent in plain English and pick from preset options.

The other thing the widget does that a chat interface doesn't: forget your previous emails. No conversation history, no model accidentally referencing the last email you wrote, no privacy concerns about what's in the chat log. Each generation is fresh.

Where AI email drafts go wrong, and how to spot it

The model has a few reliable failure modes. Watch for them before you send:

• Over-flattery in cold outreach. If the draft opens with "I've been a huge admirer of your work for years," delete that sentence. Recipients can tell the difference between specific praise ("the post you wrote about retention math last March changed how I think about cohort analysis") and generic flattery, and the AI defaults to generic unless you give it specifics.
• Hallucinated context. If you didn't tell the model what previous emails said, it sometimes invents context to bridge a gap. "Thanks for the document you sent last week" — if no document was sent, that line has to go. Always skim the draft for anything you didn't put in the input.
• Inflated formality. The model errs toward formal for ambiguous inputs. If the email is to a coworker you grab coffee with, lower the tone setting and regenerate rather than manually rewriting "I hope this email finds you well" into "hey."
• Vague closings. "I look forward to hearing your thoughts" is filler. Replace it with the specific next step you actually want: "Can you confirm by EOD Thursday?" or "Happy to grab 15 minutes if that's easier."

Privacy: what happens to what you type

The inputs are sent to a language model to generate the draft. Outputs come back to your browser; we don't store them. We don't keep a record of what you typed, who the email was about, or what the response was. If you're drafting something sensitive — a resignation, a complaint, a legal matter — read the privacy policy before pasting in proper nouns.

For most workplace email, this is a non-issue. The model isn't an HR system; it isn't going to remember that you drafted a polite decline to a recruiter. But the discipline of not pasting in genuinely confidential information (NDA-covered content, customer PII, financial figures pre-announcement) is a good one to keep regardless of which AI tool you're using.

Related tools

• Email Subject Line Generator — the subject line is what gets the email opened. If you're sending cold or to a busy inbox, run the subject through here.
• Cover Letter Generator — for the longer-form sibling: same general idea, applied to job applications.
• Paraphrasing Tool — useful when you have a draft you wrote yourself and want a second version in a different register.
• Summarizer — for the email thread you need to catch up on before replying.

Frequently asked questions

Will the recipient know I used AI?

Not if you do a final pass. Unedited AI drafts have tells — the slightly formal opener, the predictable structure, the "I hope this email finds you well." If you spend twenty seconds editing for voice (cut one filler sentence, swap one stiff phrase, add one specific detail), the result reads as a thoughtful human email. The point of the tool is to start the draft, not to send it raw.

Can I save my favorite settings?

Not currently. Each generation is a fresh start — you pick the tone and length each time. The upside is that there's no account to set up and no settings to migrate when you change browsers. The downside is that if you always write formal short emails, you'll pick those settings every time. We may add presets if enough people ask.

Does it work for languages other than English?

The widget UI is English, but you can write your intent in any language the underlying model supports — Spanish, French, German, Portuguese, Japanese, and others. The output will be in whatever language you wrote the intent in. Localized versions of the widget are on the roadmap; for now the manual route works.

How is this different from Gmail's "Smart Compose"?

Smart Compose finishes sentences you're already typing. This widget writes the whole draft from a one-line intent. Use Smart Compose for the email you mostly know how to write but want to type faster. Use this widget for the email you're avoiding because you don't know how to start it.

Does the AI know who I'm emailing?

Only what you tell it. If you put "Sarah from BigCo" in the input, it'll address Sarah and reference BigCo. It doesn't read your inbox, doesn't have access to your contacts, and doesn't know your relationship history with the recipient. The trade-off is privacy on one side and the need to provide context on the other — which is the right trade-off for an email tool.

What if the draft is completely off?

Regenerate with a more specific intent. "Email my boss" produces a generic email; "Email my boss to ask for two weeks off in July, framing it as already-discussed, casual tone" produces something you can actually send. The input is the lever. If the output keeps missing, the input needs more direction, not the model.

Does this cost anything?

AI features at Microapp run at near-cost — you pay roughly what the underlying model API costs us, plus a thin margin to keep the service running. Members get a monthly allotment of AI credits included in the membership; beyond that, you pay per use at the same near-cost rate. No subscription minimums, no annual contracts, no credit-card-on-file required to try it. Compare that to the $39–99/month subscriptions other AI email services charge, where the markup pays for the marketing budget rather than the writing.