AI Email Redactor
Bring your own API key. It is sent directly from your browser to the provider you choose — OmniDeck never sees it or stores it on any server.
How to use this AI email redactor
- Choose the situation and tone that fit your case.
- Add the specific context or details.
- Click "Draft Email" for a ready-to-send message.
Why do I need my own API key?
Running AI models costs money per request. Instead of OmniDeck paying for (and rate-limiting) everyone's usage, you use your own account's credits, and we simply provide the interface.
Is my key safe?
Your key is used only to call the provider directly from your browser. It is never transmitted to or stored on OmniDeck's servers. If you check "remember," it is saved in your browser's local storage, which only your device can read.
Getting a "blocked by CORS" error?
Some providers restrict direct browser requests for security. If a provider fails, try a different one from the dropdown — Anthropic's API is the most reliable option for this kind of direct-from-browser use.
Can I ask for a specific length?
Yes — mention it in the context field (e.g. "keep it under 100 words") and the AI will generally try to respect that.
Does this work for languages other than English?
Yes — write your context details in the language you want the email drafted in, and most providers will respond accordingly.
Why tone is harder to get right in writing than in speech
Difficult conversations are hard enough in person, where tone of voice, facial expression, and immediate back-and-forth can soften or clarify a message in real time — email strips all of that away, leaving only word choice to carry the entire emotional weight of the message. This is exactly why a message that felt reasonable while writing it in the heat of frustration can read as harsher (or more passive and unclear) than intended once it lands in someone else's inbox with none of the context of how you actually felt while typing it. Having a structured second pass — whether from a human editor or a tool like this — helps catch that gap between intended tone and how the words will actually land for someone reading them cold.
Why "firm but polite" is a genuine skill, not just word choice
The difference between an email that reads as firm and one that reads as aggressive usually comes down to a specific, learnable structure: stating facts objectively (the invoice is 30 days overdue) rather than making accusations (you're ignoring me), focusing on the specific issue rather than generalizing about the person's character, and clearly stating what you need next rather than venting about what went wrong. This pattern — closely related to structured communication techniques used in negotiation and conflict resolution training — is what separates a message that gets a cooperative response from one that puts the recipient on the defensive, and it's a skill that can be applied deliberately rather than left to how you happen to be feeling when you sit down to write.
The risk of relying entirely on AI for emotionally significant messages
For situations that matter personally — a resignation letter, a serious complaint, an apology — a fully AI-drafted message risks reading as generic or impersonal precisely in the moments where the recipient is most likely to be reading closely for sincerity. A draft from this tool is a strong starting structure and a way to get past a blank page when emotions make it hard to find the right words, but revising it in your own voice before sending — adding a specific detail only you would know, adjusting phrasing that doesn't sound like how you'd actually say it — usually produces a better result than sending the raw output unedited, especially for messages where the relationship matters more than the wording.
Why specific facts matter more than clever phrasing
The single most useful thing you can do to improve a generated draft is include specific, concrete details in the context field rather than a general description of the situation — an actual invoice number and date, the specific previous communication that went unanswered, the exact deadline being negotiated. A vague input ("client hasn't paid") forces the model to write in generalities that read as templated and easily ignored, while specific details let it write a message that sounds like it's actually about this situation, not a generic version of the category of situation — which also makes the draft more immediately usable without heavy editing.
Limitations of this tool
This tool drafts an email based on the situation, tone, and context you provide — it doesn't know your actual relationship history with the recipient beyond what you type, can't verify that any specific fact you mention (an invoice number, a date, a previous conversation) is accurate, and has no legal or HR expertise for situations where that matters (a formal resignation with contractual implications, a complaint that might escalate to a dispute). Read any draft fully before sending, correct or add anything the model couldn't have known, and treat this as a strong starting point for a message you still need to personally stand behind, not a finished, ready-to-send document without review.