Skip to content

write_example_emails

Draft sample emails for your leads

Act · mcp:actAsks the userOpens at: Target confirmed, choosing a plan30 calls / 10 minCalls a provider

Writes the emails of three leads of the preview, each from its own company, under one subject, and asks the user whether the tone sounds like them. Their “Sounds like me” hands the campaign to the autopilot: VibeSell then buys its leads with the plan's lead credits, writes each lead's emails the same way, launches it and relaunches it every month.

Description the model readsSent as is in tools/list.
Once the user said yes to their target, write the emails of three leads of their preview, each from its own company, under one subject (the follow-ups reply in the same thread), and ask the user whether the tone sounds like them. Their yes on the tone starts the campaign: VibeSell then buys its leads with the plan's credits, writes each lead's emails the same way in the background, launches it on its own and relaunches it every month; how many leads it buys is not theirs to set. It returns a question: put it to the user and call this again with their answer, never yours. After a no, pass what they want changed as guidance. Nothing is bought and nothing is sent before their yes.

The user says

Show me how my emails would read.

tools/call · Arguments the assistant sends
{
"name": "write_example_emails",
"arguments": {}
}
guidancestringoptional≤ 500 chars

What the founder asked to change in the tone of the emails, in their words.

answersobject[]optional≤ 12 items

The user's answers to the cards this tool showed, each with the id of its card. Pass back the list the last call returned, plus the new answer.

  • card stringrequired — The card id this answer was given to.
  • answer stringrequired — The choice id, or the user's own words.
Input schema
{
"type": "object",
"properties": {
"guidance": {
"description": "What the founder asked to change in the tone of the emails, in their words.",
"type": "string",
"maxLength": 500
},
"answers": {
"description": "The user's answers to the cards this tool showed, each with the id of its card. Pass back the list the last call returned, plus the new answer.",
"maxItems": 12,
"type": "array",
"items": {
"type": "object",
"properties": {
"card": {
"type": "string",
"description": "The card id this answer was given to."
},
"answer": {
"type": "string",
"description": "The choice id, or the user's own words."
}
},
"required": ["card", "answer"]
}
}
}
}

Result kindchoicesending_choicecampaign_step

Output schema
{
"type": "object",
"properties": {
"kind": {
"type": "string",
"enum": ["choice", "sending_choice", "campaign_step"],
"description": "choice: a question to put to the user. sending_choice and campaign_step: what their answer recorded."
}
},
"required": ["kind"],
"additionalProperties": {}
}
  • It returns a question with two choices, “Sounds like me” and “Change the tone”, and takes the user's own words too: the assistant asks it in the chat, as a numbered list, and passes their answer back in answers, never one of its own. Only an explicit yes counts.
  • After a no, pass what they asked to change as guidance and call it again.
  • Nothing is bought and nothing is sent before their yes. Their yes also approves the template the campaign's emails follow.
  • How many leads the autopilot then buys is not the user's to set: the assistant promises no number, no cap and no date.
  • It needs a signed-in account and a campaign in the conversation. The preview the examples are written from is taken by the tool itself when the conversation holds none.