Physical Mail API for LLMs

Designed for the model, not the dev portal

Most APIs are designed for humans reading docs. Agent Mail is designed for an LLM reading documentation cold and figuring out how to call it. Concretely:

• /llms.txt and /llms-full.txt at the root, with crisp tool descriptions.
• /openapi.json with rich descriptions and an x-mcp extension pointing back at the MCP server.
• /setup returns ready-to-paste MCP config for every major client.
• JSON-LD SoftwareApplication and WebAPI graphs on the homepage.

Tool surface

• send_letter (recipient, body, optional return address, optional subject).
• send_postcard (recipient, front image URL, message).
• get_mail_status (id).
• list_sent_mail (limit, cursor, filters).

Primitives only. No 17-knob configurations. The schemas are tight enough that an agent can call them on the first try without reading more docs.

Why this matters

If an agent has to read 8 pages of docs to figure out an API, it won't. If the surface is small, well-typed, and self-describing, it will. We optimize aggressively for the second case.

{
  "mcpServers": {
    "mailsnail": {
      "command": "npx",
      "args": [
        "-y",
        "physical-mail-mcp"
      ],
      "env": {
        "MAIL_PROVIDER": "managed",
        "MAIL_API_BASE_URL": "https://api.mailsnail.dev"
      }
    }
  }
}