You are reading the docs for Lettrove 1.0.1. Every 1.0.x release (1.0.0, 1.0.1) works as these pages describe; what each one fixed is in the changelog.
AI and your allowance
Where AI works
- In the editor, the assistant designs an email from a brief, or changes the one open from an instruction ("shorten the headline", "make the button say Reserve my seat"). Every change it makes is an ordinary edit in the email's history, with the assistant as its author, and can be undone.
- Pictures can be generated from a description, in the editor's Images panel, and land in your library like an upload.
- Reading a PDF of contacts: the model is asked only to read the table; every row then goes through the same checks as a CSV (Bringing contacts in).
- Through the MCP server, an assistant of your own (Claude, Cursor…) does the same things with its own model; see MCP server.
Which models
| Work | Model |
|---|---|
| Designing a whole email from a brief | Anthropic's Claude Opus 5 |
| Changing part of an email, the line saying what the assistant is doing, and the assistant's memory of a long conversation | Anthropic's Claude Haiku 5.5 |
| Reading a PDF of contacts | OpenAI's GPT-5.6 Luna, or Claude Haiku 5.5 when OpenAI is unavailable |
| Pictures | Image models run by fal.ai |
We move to newer models as they come out, and this page names the ones in use.
What the models see
What you give them: the brief, the email being edited, the picture description, the PDF. Never your contact lists as a whole, your sending domains, your members or your keys. An assistant connected through the MCP server sees what the tools return to it, which is what you could see yourself.
The allowance
AI use on lettrove.com, and in the editor embedded inside your products, draws from one allowance per account each period: the account that created the workspace holds it, and every workspace of that account draws from the same one. Settings → Usage shows how much of the period's allowance is used and how it divides between workspaces. When it is used up, AI features pause until it resets; everything else keeps working.
For the embedded editor, picture generation can be switched off per project under Settings → Embed → AI, so your users cannot draw from the allowance there.