List the mailboxes you can use, newest first, each with its id and email address. Call this when you need a mailbox_id for any other tool, or to check whether you already have a mailbox before creating one — it is usually the first call in a workflow. Returns every mailbox on the account, or just your own if your API key is scoped to a single mailbox. An empty list means you have none yet — use create_mailbox.

Create a NEW email mailbox this agent owns. Call this when list_mailboxes shows you have no suitable mailbox yet — mailboxes persist across sessions, so creating a second one for the same purpose strands mail in the first. Returns a working address immediately. Optional slug personalizes the address (agent-@...). To create the mailbox on a custom domain instead of the shared one (@yourdomain.com), pass a domain_id from list_domains — the domain must be verified.

List the account's custom sending domains with their id, name, and verification status. Call this when you want to create a mailbox on a custom domain (pass the id of a verified domain as create_mailbox's domain_id), or to check whether a domain has finished verifying. Only status "verified" domains can host mailboxes; an empty list means only the shared platform domain is available.

Send a new email from one of your mailboxes. Call this when starting a NEW conversation; to answer an email you received, use reply_email instead (it threads correctly). Pass a client_id (any unique string you make up) and reuse the SAME client_id if you retry after an error — that guarantees the email is sent at most once. If the mailbox requires human approval, the send is queued as a draft and returns {draft_id, status:"pending_approval"} instead of a message id — that is NOT a failure and must NOT be retried; the email goes out once a human approves it (check with list_drafts).

Reply to an email you received. Call this when answering an existing message — recipient, subject, and conversation threading are derived automatically from the original. Pass a client_id (any unique string you make up) and reuse the SAME client_id if you retry after an error — that guarantees the reply is sent at most once. If the mailbox requires human approval, the reply is queued as a draft and returns {draft_id, status:"pending_approval"} instead of a message id — that is NOT a failure and must NOT be retried; it is sent once a human approves it (check with list_drafts).

List recent messages in a mailbox, newest first, with comprehension status per message. Call this when checking what has arrived; use read_email for full content and extracted data.

Read one email in full: cleaned body, sanitization report (screened against known hidden-content patterns), classification, and data extracted to the mailbox schema. Call this when you need the content or extraction of a specific message.

Block until a matching email arrives in a mailbox (or time out). Call this when you just sent an email and need the reply, or are expecting an inbound message — instead of polling list_inbox yourself. Returns the full message on match, or {timed_out:true}.

Prepare an email as a draft without sending it. Call this when you want to stage a message for later or for a human to review before it goes out — send it afterward with send_draft. Same fields as send_email, or omit to/subject and pass reply_to_message_id to draft a threaded reply. Returns the draft id with status "draft".

List the open drafts in a mailbox (status draft, pending_approval, or sending), newest first. Call this when you need to see messages you have staged or that are queued waiting on human approval, e.g. after a send returned status "pending_approval".

Read one draft in full by its id: recipient, subject, body text, and current status. Call this when you need the content of a specific draft — e.g. to review what is queued for approval after list_drafts, or before editing it with update_draft.

Edit the to, subject, or text of a draft that has not been sent yet — omitted fields keep their current value. Call this when a draft needs changes, e.g. a human declined to approve it and you are revising it. The reply target cannot be changed (create a new draft to reply to a different message), and a draft that is already sending or sent can no longer be edited.

Send a draft you prepared, by its id. Call this when a staged draft is ready to go out. If the mailbox requires human approval, this returns 202 {draft_id, status:"pending_approval"} — the draft is queued for a human to approve, which is NOT a failure and must NOT be retried; it is sent once approved. Otherwise it sends immediately and returns the message id.

Discard a draft by its id so it will never be sent. Call this when a staged or pending-approval message should be withdrawn — e.g. it is no longer needed or was created by mistake. A draft that is already sending or sent cannot be deleted.