FAQ
Frequently asked questions
By Justin Winter · Updated September 17, 2026
paperplane prints and mails real USPS letters and postcards from $1.99, with a REST API and MCP server for AI agents. Below are straight answers to the questions people ask first — what we're built on, how content gets screened, who holds your money, what Certified Mail actually proves, and how long we keep your data.
This page collects the questions we expect first from a technical or skeptical audience — the kind that shows up when a product like this gets shared somewhere like Hacker News. Each answer links to the fuller page or the actual code/policy behind it, so you can go deeper rather than take our word for it.
Still have a question?
Most operational and USPS-mechanics questions (how long delivery takes, how tracking works, what happens to undeliverable mail) are covered in our guides. For anything about security posture and infrastructure specifically, see Trust & security. Anything else — reach out through the contact page.
Related guides
Common questions
Why build this instead of using Lob or PostGrid?
Lob and PostGrid are mail-rail infrastructure — APIs a service like ours could route print jobs through, not consumer-facing products a person picks between. We are not competing with them; a version of paperplane could run on top of one of them. Our comparisons (/compare) are against other consumer-facing letter-mailing services — Mailform, Click2Mail — for that reason.
How does content screening work, and what stops abuse?
Every order runs through an automated content screen before anything is printed. In production, screening runs on a real model and fails closed: if no screening provider is configured, orders hold rather than silently skip the check — there is no silent bypass path. Lawful content (demand letters, firm legal language, blunt personal letters) is explicitly allowed; what gets refused is threats, harassment, and impersonation. Screening verdicts (categories, not the letter text itself) are retained for audit and legal compliance.
Who holds my money — do you custody funds before a letter ships?
Stripe does, and only briefly. We use Stripe's manual-capture mode: your card is authorized when you place an order, but the charge is only captured after the letter is actually screened and queued for print — if screening refuses an order, the authorization is simply released, nothing is captured. We never hold a stored balance or wallet on your behalf outside of purchased mailing credits, which are visible in your account.
What does "Certified Mail" actually prove, legally?
Certified Mail gives you a USPS tracking number and a mailing receipt — proof that something was sent, on a specific date. It does not by itself prove what was inside the envelope, and the optional electronic Return Receipt adds a signed delivery record, not a guarantee that mailing satisfies your specific legal notice requirement. Court and statutory service rules vary by state and by matter — see our guide on proof of service before relying on any mailed notice for something with a deadline attached.
How long do you keep my documents and data?
Your uploaded document is stored encrypted and deleted 30 days after delivery — we don't keep a permanent copy. Addresses stay in the order record, and the automated screening verdict (categories, not your letter text) is retained for audit and legal compliance. Payment details are never seen or stored by us at all; Stripe handles that entirely. Full detail on /privacy.
Is this a real, working product or a demo?
It's real: the send pipeline, pricing, screening, and tracking are live code, not mockups — try it end-to-end for free at /demo with sandbox mode, which runs the actual pipeline (address checks, content screening, simulated printing) without mailing anything or charging a card.
What happens if I paste something that gets refused by the content screen?
You get a clear reason and your payment authorization is released — nothing is charged. Refusal categories are things like threats, harassment, and impersonation, not disagreement with the content or tone of a lawful letter. If you think a refusal was a mistake, contact support; screening decisions are logged, not black-boxed.
Can an AI agent send mail on my behalf without me approving it?
Not in one step. The MCP and REST send tools require a confirmation_token minted by a prior, separate quote call, bound to the exact recipient, content, class, and price — an agent cannot invent a send from nothing. A live send still returns a Stripe payment link for a human to approve before any card is charged; sandbox mode skips payment entirely and mails nothing, which is the safe way to let an agent explore.