Templated email that reads like a person wrote it.
The humanizer is off unless you turn it on. When it's on, a model rewrites each templated email in the tone you pick, just before it's sent.
Three tones, where you want them.
Casual, professional or friendly, with professional as the default. Turn it on for the whole account or for one agent, and set a different style for one mailbox. Any single send can still be delivered exactly as written, even when the account has it on.
{ "tenantId": "…", "recipientEmail": "priya@acme.com", "templateId": "renewal", "dedupeKey": "renewal-priya-1", "humanize": false}Only the prose changes.
The model is told to keep the HTML, links, unsubscribe links and compliance text exactly as they are, and not to add greetings or sign-offs. The template's own output is stored next to each rewrite.
casual · professional (default) · friendly · Anthropic or OpenAI
See the rewrite before it's sent.
Preview any email from the CLI, MCP or API and get the setting that applies, the template's output next to the rewrite, a line diff and the result of the checks. Nothing is sent. Each preview that reaches the model uses one rewrite from your plan, whether or not the rewrite passes the checks. Like it? Send with its preview id and exactly that rewrite goes out, with no second rewrite and no second charge. A send held for approval shows the approver the rewrite, and that is what's delivered.
molted humanizer preview \ --template renewal \ --payload '{"firstName": "Priya"}' \ --diff molted send --to priya@acme.com \ --template renewal --dedupe-key renewal-priya-1 \ --payload '{"firstName": "Priya"}' \ --humanize-preview <previewId>Hand it to your agent.
It sets up its own inbox. You decide how much it does on its own.