openemail tracking

Returns one page of the messages the workspace tracked and sent in the window, newest first, each as a full engagement report with its recipients and links. It covers the whole mailbox, not only mail sent through this API: messages from the composer, the MCP tools and the assistant appear too, with sendId null where no send record exists. A report built off emails.list would describe the API rather than the mailbox.

Messages that carried no pixel and no rewritten link are absent entirely and never appear as a zero. opened: false therefore means tracked and not opened, and clicked narrows the same way. Both filters work on counted opens and clicks, so a message fetched only by Apple Mail Privacy Protection or a link scanner still counts as unopened.

The window defaults to 30 days. minutes wins over days when both are set, and the start of the window is floored to the grain in UTC, so days: 7 covers today plus the six whole days before it. Paging is keyset: nextCursor is the tmsg_ id of the last report on the page, and passing it back as cursor with the same filters continues strictly after it, so every tracked message in the window is reachable.

Add --all to walk every page: a table on a terminal, one JSON object per line when piped or with --ndjson, and one { items, hasMore, nextCursor } document with --json. --max <n> stops after that many items.

Returns the numbers behind an engagement panel in one request: how many messages were tracked, opened and clicked, the open and click rates, hit totals, a time series, the top links, mail clients and countries. It covers every tracked message sent from the mailbox in the window, whichever surface sent it.

Rates and totals count different things, and mixing them is how open rates above 100% get published. opened and clicked count distinct messages. totalOpens and totalClicks count hits. openRate is a percentage of trackedForOpens, the messages that actually carried a pixel, and clickRate is a percentage of trackedForClicks, the messages that had a link to rewrite, which is usually far fewer than tracked because most replies contain no links.

byDay is sparse: a bucket in which nothing was sent has no entry, so a chart must fill the gaps. grain sets the bucket width and the key shape, YYYY-MM-DD, YYYY-MM-DDTHH or YYYY-MM-DDTHH:MM, and --offset-minutes shifts bucket boundaries so days break where the reader's day does. The window starts at the beginning of the oldest bucket, so that bucket is whole, and ends now, so the newest is partial.

Returns the whole report for one tracked message: totals, one entry per tracked copy under recipients, and every rewritten link with its clicks under links. It accepts either identifier a caller may hold. A msg_ id is looked up as the send it went out as, and anything else is treated as a tmsg_ tracking id, which is what tracking.list and webhook payloads carry for mail sent outside this API.

openCount and clickCount are counted readings: opens that looked like a person, with repeat fetches within thirty seconds collapsed. openCountRaw and clickCountRaw include every hit, machine fetches such as Apple Mail Privacy Protection and corporate scanners among them. Gmail's image proxy is a real person displaying the message and is counted, but only once, because later views are served from its cache.

recipients is one entry per tracked copy, which is not always one entry per person. When one body went to the whole list, or the bytes could not vary per recipient, the shared copy has email null and attributed false. Branch on attributed and on the message-level attributable before saying that a named person has or has not read anything.

Returns one page of the raw open hits for one tracked message, newest first. Each hit says which copy was fetched, how it was classified, whether it moved the numbers, and what the User-Agent and the edge could tell about the client, device and location.

kind is human for something that looked like a person reading, proxy for an image proxy such as Gmail's, which is a genuine reading by a reader the tracker cannot see, and machine for a scanner or Apple Mail Privacy Protection, which fetches on delivery and only proves the message arrived. By default only counted hits come back. includeMachine: true adds the hits that did not count, which is every machine hit plus repeat fetches within thirty seconds of a counted one on the same copy, so it filters on counted rather than on kind.

A msg_ send id is resolved first, and one that names no tracked message is a 404 rather than an empty list, since an empty list reads as nobody having opened it. A tmsg_ id is taken as given without a lookup, so one that never existed or belongs to another workspace comes back as an empty list. Only read an empty result as an answer about readers when the id came from this API.

Add --all to walk every page: a table on a terminal, one JSON object per line when piped or with --ndjson, and one { items, hasMore, nextCursor } document with --json. --max <n> stops after that many items.

Returns one page of the raw click hits for one tracked message, newest first. Each hit has the same classification and client detail as an open, plus linkId and url naming the link that was followed, where url is the original destination rather than the redirect.

A click is stronger evidence of reading than an open. Mail clients block images far more often than readers leave links unfollowed, so a message with clicks and no opens was certainly read. Link scanners that follow every URL on delivery are classified as machine and excluded from the default result, and includeMachine: true brings them back along with repeat hits collapsed by the thirty second window.

Only links in the new part of the body were rewritten, so nothing here can be a click on quoted history under a reply. The id rules match listOpens: a msg_ id that names no tracked message is a 404, while a tmsg_ id is used without a lookup and an unknown one returns an empty list.

Add --all to walk every page: a table on a terminal, one JSON object per line when piped or with --ndjson, and one { items, hasMore, nextCursor } document with --json. --max <n> stops after that many items.