Getting Started

Welcome to XPT Tracker — one workspace for tasks, projects, documentation, and AI agents. This guide walks you through the essentials, from your first task to the knowledge base and working with AI agents.

How AI works in XPT Tracker

AI agents are first-class participants here, not a separate chat box. Three kinds work alongside you:

Not sure which one fits? Which kind of agent to use compares them side by side.

All three act with the same, permission-checked access as a person. They can:

• read and update tasks, projects, and milestones;
• take part in discussions;
• write and maintain documentation across the knowledge base.

Local agents go one step further: they plan and implement tracker tasks end to end — read a task, write a plan, do the work in your codebase, run the tests, then record the result and learnings back in the workspace. They follow your workspace's AI Workflows: the same plan → implement process, defined once and followed by every agent.

Every action is recorded in the activity history, attributed to the person who triggered it. The AI Agents and Work from Your Own AI (MCP) sections go deeper.

Signing in

Open your workspace URL and sign in. You land inside your workspace, with the navigation sidebar on the left and your content in the main area. If you belong to more than one workspace, use the workspace switcher at the top of the sidebar to move between them.

Finding your way around

The bottom of the sidebar also holds your avatar menu (profile, invite, log out) and the light/dark theme toggle.

Your first task

1. Open a team's board from the sidebar.
2. Click + in the header of a status column. In the New task form give the task a title and press Create task.
3. Open the task to fill in details: description, assignee, priority, labels, an estimate, and relations (subtasks, dependencies, project, milestone).
4. Drag the card between columns as the work progresses. Everyone watching the board sees the change in real time.

That's the loop. Boards & Statuses covers the board, and The Task in Detail explains every field and action.

Bringing in an agent

Anywhere you can write a message — a task's Discussion tab, a brainstorm, a channel — type @ to pick an agent and ask it to do something: summarize a thread, draft a subtask breakdown, create tasks from a discussion, or look something up in the knowledge base. The agent runs, may call tools, and posts its answer back into the conversation. Using Agents shows the ways to talk to one.

When the job needs files and real computation — joining several spreadsheets, reading a PDF report, running a calculation, collecting data from public sources — mention a harness agent. It works in its own sandbox, keeps its files between your messages, and returns the result as a file attached to its reply. See Harness Agents.

To add an agent to the workspace, see Configuring an Agent. For work that comes back — watching sites, a weekly digest, a summary kept fresh — give an agent an AI Job and it runs on a schedule by itself.

If you'd rather drive the workspace from your editor or terminal, connect a local agent over MCP: Connect Claude, Connect ChatGPT and Codex, or any other MCP client.

Where to go next