Agent Direct Mail

What direct mail looks like when an AI agent is the marketer: targeted, one-piece-at-a-time, decided per-event. Patterns, examples, and how to set it up.

use-casedirect-mailmarketing

· Updated

A different shape

Classic direct mail: pick a list, pick a creative, blast 10,000 pieces, hope 1% respond. Agent direct mail: agent watches for high-intent signals, drafts a piece per signal, mails one. Volume drops 100x; relevance jumps 100x.

What a setup looks like

  1. Agent monitors signal sources (your CRM, product analytics, social listening).
  2. When a high-value signal fires (a target account hit your pricing page three times this week), the agent enriches the lead, drafts a tailored letter, and calls send_letter.
  3. get_letter updates your CRM when the piece is delivered.

Why this works

Physical mail, sent at low volume but high relevance, has open rates that crush email and a tactile quality that emails never get. The bottleneck has historically been the human time required to draft and send each piece. The agent removes the bottleneck.

Caveats

  • The bar for content quality goes up. A bad personalized letter is worse than a generic one.
  • Set hard rate limits per recipient. Keep a log of each send's receipt_token and check it before mailing again.
  • Track response attribution carefully — physical mail responses often happen days later, in inbound channels you wouldn't otherwise instrument.

Wire Mailsnail into your agent

Add one URL and sign in. Nothing to install and no API key. In Claude Code:

Claude Code
claude mcp add -t http mailsnail https://api.mailsnail.dev/mcp

Then run /mcp and choose Authenticate. In Claude Desktop or ChatGPT, add https://api.mailsnail.dev/mcp as a custom connector. Cursor and other JSON-config clients take:

mcp.json
{
  "mcpServers": {
    "mailsnail": {
      "url": "https://api.mailsnail.dev/mcp"
    }
  }
}

A one-page letter is from $1.21, and preview_letter shows the proof and exact price before anything is charged. Setup for every app →

See also