ChangeEngine MCP: what's live, what's next

The communications inside one program, last edited first, each with its status, channel and who touched it last. Archived ones are left out.

The communications and surveys the account changed most recently, newest first. The fastest way to answer what is the team working on right now.

Search past communications by intent and get back resource links.

The order a person meets an account's communications, grouped by the date each one hangs off. This is how you see the shape of a journey, such as everything a new starter receives and when. Each line reads back in the same words set_schedule takes.

The whole review trail for one communication: each step, who took it, and the note. A changes-requested note is the answer to why something is not live yet.

Create an empty communication draft inside a program.

Rename a draft, rewrite its internal description, or move it into a different program, without touching its copy, audience or schedule.

Target a named segment, an ad hoc segment expression composed from existing condition fields, a manager or manager's-manager expansion of a segment, or a Slack/Teams channel directly.

Override who the communication comes from, so a message can read as the People Team or a named person rather than the app. Pass null to fall back to the account default.

Every active segment, with a flag for which can act as a base segment.

The employee fields that can pick a group out of a base segment, with how many of its people carry each field and the values it holds. Call it before set_audience to build an ad hoc segment on a field or value no existing segment uses yet. Fields only a few people carry, and fields whose values are nearly all unique, are left out. A special category field such as gender or religion reports its coverage without its values.

The dates an account can schedule against and when each one fires. No longer only birthday, start date and work anniversary: whatever the account actually runs, including became a manager and became an admin.

The uploaded files and Design Huddle templates in the account's content library, grouped by folder and flagged by type. Use approved artwork in a message instead of generating something new. Uploaded files come back with a public URL, documents such as PDFs are flagged so you link them rather than embed them, and a template comes back with a Template ID for render_content_library_template.

Prepare a Design Huddle template from the content library for a message. It returns the design node to drop straight into set_message plus a preview image of what that node looks like. The design stays editable in the app afterwards, so a branded template can be used in an agent-built message without flattening it into a picture.

The senders available to set_sender for a communication, and which one it currently uses.

See a communication rendered as its recipients will, with sample merge values, returned as an image. A tall message comes back as several images, top to bottom. This is the answer to a write that succeeds and still looks wrong.

Add a page to the draft.

Different completion text depending on how someone answered.

Save the completed translation set back onto the survey.

Stop a live survey taking answers, with your own message on the closed link. Answers already given stay, and launch_survey reopens it.

Non-archived surveys with status, response counts, and live links.

What people actually answered. Counts per answer for closed questions, open text tallied by theme, then whole responses so one person's rating and their comment read together. No answer is tied to a person and no identifier comes back; an anonymous survey shows nothing until 5 people have answered.

The same numbers for one individual send, most recent first, for the account, one program or one communication. Adds the channel, the segment, who published it, the trigger that fired it, clicks broken down by URL, errors broken down by reason, and the name and address of anyone it failed to reach. Everyone who opened, clicked or reacted also comes back by name, most engaged first, with the URLs each person clicked and how many times; a send names up to 50 of them and reports how many engaged people are not in the list. A hard bounce comes back with the full SMTP reason, so a bad address is finally actionable.

What people answered in the polls a message asks. Each poll comes back with its question, how many answers it has, and the count and share behind every option, including an option the poll no longer offers if somebody chose it, so the shares still add up. No answer is tied to a person. A message with no poll block returns nothing, and a poll nobody has answered yet says so rather than reporting zero as a result.