Skip to main content
POST
Every send is queued — we accept the request, classify the recipient, pin (or race) a backing mailbox, and the dispatcher delivers from there. The response carries a pendingId you can correlate with the downstream email.sent webhook (or email.send_failed_permanently on a hard fail). The body has two shapes — new conversation or reply — that are mutually exclusive. New conversations need to and subject; replies need convId and we resolve recipient, subject (with Re:), threading headers, and the previously-pinned mailbox from the existing thread. text and html can both be set; recipients see the multipart message their client prefers. Most production senders set both.

​
Send classification

The response includes sendClass, which tells you how we routed:
  • cold_first_contact — first message ever from this identity to this recipient. We pace these with the identity’s drip schedule and honor the working-hours window. The first available mailbox to claim the row owns the recipient from then on; subsequent sends use the same mailbox.
  • cold_followup — second-or-later cold message to a recipient who hasn’t replied in the last 3 outbounds. Same drip + working hours as a first contact, but pinned to a known mailbox.
  • warm — replying to a recipient who responded within their last 3 outbounds. Bypasses drip pacing and the working-hours window — responsive recipients get answered as fast as the dispatcher fires.
pinnedAccountId is the backing mailbox id (only meaningful for cold_followup and warm; null on cold_first_contact until a mailbox claims the row).

​
Labels

Label the conversation at queue time by passing labels: [...]. Idempotent on the conv (re-labeling an already-labeled conv is a no-op), so sequencer flows can pass the same campaign label on every step.
The per-label rollup analytics (sent / replied / positive / bounced) pivot off these labels — see GET /identities/:id/labels/stats. Free-form strings; we recommend a key:value convention so labels filter cleanly on prefix.

​
No-reply nudges

Pass noReplyEventAfter (duration string "1d" / "4h" / "30m" or a raw number of milliseconds) to schedule an email.no_reply webhook + event when no inbound lands on the conversation within that window. Default: 1 day. Floor: 60 seconds.

​
Idempotency

Pass Idempotency-Key to make retries safe. Same key with the same body returns the original response (header Idempotent-Replayed: true); same key with a different body returns 409. Keys live for 24 hours, scoped per-org. For sequencers the right key is <leadId>:<step> — if your worker crashes after we accepted the send but before your DB recorded it, the retry returns the same response and you don’t double-send.

​
Stitching to external threads

If the conversation started outside 12m (an imported CRM history, forwarded mail), pass inReplyTo and references on the new-conv shape. We thread the new send onto the external chain.

​
Status codes

  • 202 Accepted — at least one row was queued. The body’s results[] array tells you per-recipient outcome.
  • 429 Too Many Requests — every recipient on the request was rejected (do_not_contact, no_eligible_pool, inactive, suspended). Inspect results[i].reason to discriminate.
  • 409 Conflict — Idempotency-Key reused with a different request body.

​
What we don’t support yet

  • Scheduled sends. No sendAt. We pace cold sends via the identity scheduling config (working hours, drip interval); for one-shot future sends, run a scheduler in your own infra and call at fire time.
  • Attachments. Not supported.
  • Custom headers. No way to set List-Unsubscribe / List-Unsubscribe-Post today. High-priority because of bulk-sender requirements; we’ll ship it.

Authorizations

​
Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Headers

​
Idempotency-Key
string

Up to 255 characters. Same key with same body returns the original response (with Idempotent-Replayed: true); same key with a different body returns 409. Keys live 24 hours, scoped per-org.

Maximum string length: 255

Path Parameters

​
handle
string
required

Identity handle, URL-encoded.

Example:

"alice.acme@inboxbase.ai"

Body

application/json
​
to
required

Recipient address. Pass an array to send to multiple recipients; each becomes its own conversation.

​
subject
string
required

Subject line.

Minimum string length: 1
​
text
string

Plain-text body. At least one of text / html is required.

​
html
string

HTML body.

​
inReplyTo
string

Message-ID of an external message you're stitching to.

​
references
string[]

Cumulative References chain for stitching to an external thread.

​
noReplyEventAfter
default:1d

A duration string ("4h", "3 days", "90 minutes", "1.5h") or a raw number interpreted as milliseconds. Floor 1 minute.

Example:

"4h"

​
labels
string[]

Conversation tags applied at queue time (e.g. ["campaign:Q4-launch", "priority:hot"]). Idempotent — a tag already on the conv is a no-op. Pivot label-rollup analytics off these via /identities/:id/labels/stats.

Maximum array length: 24
Required string length: 1 - 120

Response

At least one recipient accepted.

Queue-first send response. The dispatcher delivers each row asynchronously; correlate pendingId with the downstream email.sent webhook to know when delivery actually fired.

​
status
enum<string>
required

queued if at least one recipient queued; rejected if the whole batch was rejected.

Available options:
queued,
rejected
​
identity
string
required

Identity handle echoed back.

​
queued
integer
required

Count of recipients successfully queued.

​
rejected
integer
required

Count of recipients rejected.

​
results
object[]
required