[For agents and agent builders]

Give AI agents postal mail sending, drafting, and receiving workflows

Send certified mail, letters, notices, and postcards with tracking, webhooks, and proof. Turn forwarded files from an address you already use into digital/OCR context today. New mailbox.bot-issued physical mailing and package addresses: Not live yet · Launch ETA: mid/late September. Reservations are open.

REST
/api/v1
MCP
/api/mcp
Discovery
llms.txt
Agent address manifest

reachable by API, email thread, webhook, MCP, and postal mail

postal: certified allowed
email: AgentMail thread
mcp: /api/mcp
proof: webhook + photos
Scanning outbound postal mail proof
dry_run
approval
delivery_proof
The point is not another SDK. The point is giving agents a real-world action layer with permission checks and proof.
[Agent address paths]

Use the live tools now. Join the managed-address reservation list when the agent needs one.

Outbound mail and forwarded digital/OCR context from an existing address are live. Mailbox.bot-issued receiving addresses: Not live yet · Launch ETA: mid/late September. Reservations are open, but they do not assign or activate an address today.

[Machine path]

Expose enough structure for tools and LLMs to find the right doorway.

Agents can discover auth through auth.md and OAuth protected-resource metadata, then follow stable summaries, cards, schemas, and tool catalogs.

[Featured integrations]

The strongest distribution wedge is agent-native tools that already own the digital context.

Goose and AgentMail show the category better than a generic SDK page: one starts from local SMB operations, the other starts from agent email threads.

[Live surfaces]

Public endpoints agents can discover, inspect, and call.

This is the practical part: stable URLs, machine-readable schemas, MCP tools, and webhookable actions.

[Workflow proof]

Concrete loops beat feature lists.

The pages below are useful for humans and for agentic search: they map postal mail to specific buyer workflows.

[Operating model]

Let agents move the work, keep humans on the checkpoints.

That is the stance: useful automation without pretending identity, money, legal risk, or certified mail approvals vanished.

01
Agent creates or discovers the workflow.
02
Human completes identity, payment, and sensitive approvals.
03
mailbox.bot sends mail, captures inbound context, and emits proof.
04
The agent updates the original system when mail is pending, mailed, delivered, failed, or returned.
POST/api/v1/signupagent-started
curl -X POST https://mailbox.bot/api/v1/signup \
  -H "Content-Type: application/json" \
  -d '{
    "full_name": "Jane Smith",
    "email": "operator@example.com",
    "password": "securepassword123"
  }'

# Next:
# human verifies identity + payment
# agent creates sk_agent_test_
# workflow starts with dry_run=true