What it does not do

• A team coordinating people. A task can name the person accountable for it, and your inbox holds what waits on you, but there are no sprints, no story points, no velocity, and none of it is planned. If you want a tracker that coordinates humans, buy one of those.
• Anyone wanting a Jira replacement. Nothing syncs back. A sources reference points at the client's ticket and stops there, so the client's tracker stays the client's.
• A team whose tasks are written by hand. Four required slots and a criterion per task is a cost a person pays on every ticket and an agent pays in the same turn.
• A team that wants a different schema per client. The slot vocabulary is the same in every workspace, so a task means the same thing wherever it was written.

Each of these is a choice, and saying so is a position.

• Sprints, cycles, velocity, story points, burndown, capacity planning.
• No notifications sent to a person: no email digest, no chat message, no push. What waits on you sits in the inbox inside your workspace until you open it. Signed webhooks exist so your own integration can react to events; what it does with them isn't trackxai's job. A workspace admin registers one endpoint per integration in Settings, and trackxai tries a delivery up to three times over about a minute before leaving the change for your integration's own catch-up read.
• Two-way sync with another tracker.
• Rich text and embedded media. Markdown in a slot, and a link. A task, an epic or a document can carry an attachment: an image or a PDF, text, CSV, Markdown, JSON or zip file, up to 2 MB when an agent attaches it and up to 10 MB when you attach it in the browser.
• A chat interface. Agents talk to it through tools, people through the screen.
• A model judging substance on the write path. A model call per write is a cost and a delay on a path with a sub-second budget.
• A configurable schema. Every configurable vocabulary becomes an empty one.