Get a mailing address
for your AI agent

Inbound + outbound postal mail for AI agents

Outbound: Mission critical business postal mail as a service - have your AI agent send us outbound mail orders through the API & MCP. OpenClaw, Cursor, Devin, Codex, Hermes, or your IDE of choice: quickstart copy and paste the /api-docs and /mcp-install documentation in your agent ide or cli and get it up to speed ASAP to start sending mail. All document pages are printed with our high-quality, high-resolution equipment. Our team takes a picture of every single mail item on the outside of the envelope with the enclosed documents visible, tracked, and logged for your agent to reference with carrier tracking numbers. Nothing slips through the cracks ever. You can search all sent mail orders by human UX in the /dashboard or have your agent search through the API if you ever need to pull something up by tracking number, date, or sent to address.

Inbound: mailbox.bot enables an AI agent to be associated with a physical mailing address and reached in the real world. It is an early step toward giving an agent practical form and a shareable identity surface, unlocking trust and scalability. The result is a place humans, businesses, and other agents can interact with through mail, documents, notices, and eligible packages. The PMB number is assigned after USPS Form 1583 verification.

Today, mailbox.bot supports outbound physical mail as a service with scans, updates, and iron clad tracking. Agents and software can send letters, certified mail, document packets, notices, postcards, and batch mail through the dashboard, REST API, MCP, A2A, and webhooks. Agents can also get a real street mailing address with a mailbox number at our staffed Manhattan Beach, California facility. USPS requires Form 1583, signed before a notary with two forms of ID, so a verified human stands behind every agent mailbox; the notary session is online and the fee is waived. Get a mailing address # from signup.

Most teams spend money at the paper edge: checking mailboxes, opening envelopes, scanning pages, naming PDFs, copying details into a CRM, emailing status updates, printing replies, buying postage, and chasing proof. mailbox.bot turns that paper surface area into workflow context. Physical mail, packages, email, phone, CRM, ticketing, and correspondence systems can connect through scans, OCR, extracted fields, MCP/API replies, webhooks, approvals, tracking, proof, and audit events, so the full chain becomes observable and traceable instead of stitched together by hand.

mailbox.bot bridges the paper surface area of the world into agent-readable context. An existing address can feed scans, photos, PDFs, OCR, extracted fields, and notes into the workflow, and a mailbox.bot-issued address adds managed arrival photos and eligible package events to the same loop. The agent or workflow can use that context to draft the next email, phone follow-up, CRM update, support ticket, legal response, certified letter, package action, or outbound postal reply.

Yes. mailbox.bot is designed so physical street addresses, receiving workflows, outbound mail, and packages can wire into the same systems that already manage email, phone calls, CRM records, tickets, documents, and correspondence. MCP/API tools handle the agent actions; webhooks, approvals, sandbox keys, tracking, delivery proof, photos, and audit events make the physical chain observable and traceable. Outbound mail and existing-address inbound context work today; mailbox.bot-issued physical receiving addresses are at the Manhattan Beach, CA facility.

The agent can classify the mail, extract entities and deadlines, connect it to a CRM or ticket, draft a response, request approval, send outbound mail, and preserve the full lineage from inbound scan to outbound proof and webhook events.

Yes. mailbox.bot is harness-agnostic. OpenClaw, Hermes, LangChain, AutoGen, CrewAI, Goose, OpenAI Agents SDK, Vercel AI SDK, LlamaIndex, custom workers, and other AI agent harnesses can use mailbox.bot through MCP, REST API, A2A, OpenClaw discovery, and webhooks. Also your IDE such as Cursor or Devin work seamlessly. The preview name "My Hermes Agent" is an example, not a requirement. Outbound physical mail and existing-address inbound context work today.

MCP server lets AI agents send physical mail. With mailbox.bot MCP tools, an agent can prepare outbound letters and certified mail, run sandbox tests, apply human approval gates, receive webhook events, and use inbound scans or PDFs as context for replies.

The agent can add mailbox.bot as an MCP server, call the REST API, use A2A, or trigger a dashboard-approved workflow. It submits the document, recipient address, mail class, metadata, cost controls, and approval policy. mailbox.bot then handles print, postage, mail handoff, tracking, and webhook events.

It can. Use requires_approval=true, key-level force approval, sandbox keys, dry runs, and max-cost headers to keep agents from sending or spending without the operator's policy. Sensitive mail can pause for human review before live fulfillment.

No. You can manage the workflow from the dashboard, then add API calls or webhooks when you are ready. Teams that want automation can connect immediately, and teams that prefer a UI can start there and grow into the API later.

Mail documents are encrypted at rest and regularly purged. Remaining proof and lineage are retained as rasterized records for security.