ALLYX ONE
100% browser-based · zero uploads

AI Email Writer

Draft any email in seconds — cold outreach, professional replies, casual messages, or follow-ups. Free, no account required, and nothing leaves your device. Describe what the email is about and the AI writes it right in your browser.

Email type

0 / 500

Generated email

How It Works

1

Pick your email type

Choose from Cold Outreach, Professional Reply, Informal, or Follow-Up. Nothing is sent anywhere — your input stays entirely in browser memory.

2

Describe your email

Enter what the email is about and any key points to include. Optionally add the recipient's name and your own to personalise the output.

3

AI drafts in-browser

SmolLM2 360M runs inside a Web Worker using WebGPU (WebAssembly fallback). Your context and details stay entirely on your device.

4

Edit, copy, or download

Review and tweak the draft in the output field. Copy to clipboard or download as a .txt file — all in the browser.

Email Types

Type Description Best For
Cold Outreach Persuasive first contact — concise and direct Sales, partnerships, networking with new contacts
Professional Reply Formal, courteous response with clear structure Client updates, job applications, business enquiries
Informal Warm, casual tone — feels natural and friendly Colleagues, friends, casual check-ins
Follow-Up Polite check-in that moves things forward Post-meeting, unanswered emails, pending proposals

Frequently Asked Questions

Is my information uploaded to your servers?
No. Everything you type stays on your device. The AI model runs inside your browser using Transformers.js and generates the email locally. Nothing is transmitted to or stored on any server.
Why is the first generation slow?
The first time you generate an email, your browser downloads the SmolLM2 360M model (~388 MB) and saves it locally. After that one-time download, the model loads instantly from your browser's storage and works even offline.
Which AI model powers this email writer?
The tool uses SmolLM2 360M Instruct by HuggingFace — a compact, instruction-tuned language model that runs as a 4-bit quantized ONNX model via Transformers.js inside a Web Worker, accelerated by WebGPU where available. The one-time download is approximately 388 MB.
Can the AI invent details I did not provide?
Yes. Like all language models, SmolLM2 360M can occasionally infer or embellish details not in your input. Always review the generated email and correct anything inaccurate before sending — the output field is fully editable.
Does this work offline?
Yes. After the model is downloaded on your first use, it is saved to your browser's local storage via IndexedDB. You can then write emails without an internet connection.