Getting started

Your workspace is set up and run by AiM Infosoft. There is nothing to install. To get in, ask your administrator for an account; there is no public sign-up.

This page follows the path from your first sign-in to your first approved change. Each step links to the page that explains it in more depth, so you can follow the short path now and come back for the reasons later.

The order matters a little. Agents read the project's goal, notes and outcomes every time they work, so writing those before the first ticket gives the first agent something to work from.

#1. Sign in

Sign in with the details your administrator gave you. What you see in the menu depends on your group: an administrator decides which screens each group can use.

#2. Turn on two-factor sign-in (optional)

From your account menu, open the security page and enrol an authenticator app. You can choose to trust a browser for a while so it does not ask for a code every time, and you can remove trusted browsers from the same page.

#3. Create a client

Under Clients, add the organisation the work is for. Clients hold projects, so this comes first.

#4. Create a project

Under Projects, create a project for the client. Your administrator will usually have connected your code host already, so you can pick the repository from a list. Set the Agent field to the agent new tickets should go to when nobody picks one.

#5. Write the goal and a few notes

On the project form, fill in the Goal: the end state this phase is working toward, written so someone could check it is true. Then open the Specs & notes tab and add a handful of notes:

• a context note with the background a newcomer would need,
• a constraint for anything that must stay true,
• a guidance note for how you prefer things done,
• a decision for a choice already made, and why.

Every agent reads all of them. See Goals and notes.

#6. Add outcomes and run the checks

Open the Outcomes tab and add the lasting capabilities the system needs to reach the goal. Three to seven is typical for a phase. Tick each of the six quality checks as the outcome honestly passes it. An outcome can only move to Ready once all six are ticked.

Press Review with AI for a second opinion on all of them at once. It never ticks a box for you. See Outcomes and quality checks.

#7. File a ticket

Create a task in the project. Give it a clear title and a description (the form reminds you that this is what the agent reads), and link it to one Outcome. Attach any screenshots or PDFs the work depends on.

#8. Draft acceptance criteria

On the ticket, press Draft With AI in the acceptance criteria panel. It drafts two lists from the title and description: the plain criteria the client signs off, and the technical criteria the reviewer checks. Edit them until they are true. See Tickets and acceptance criteria.

#9. Assign an agent

Press Assign to agent on the ticket. Pick the agent and the execution mode (start with Implement in branch). Add a one-line instruction if there is something the ticket does not say, then press Queue the run.

The run screen shows what the agent is doing as it works. If it needs something it was not given, it stops and asks rather than guessing.

#10. Review and approve

When the run reaches Awaiting approval, open it. Read the changes, look at the screenshot, read the agent's report, and check the work against the acceptance criteria. Then choose:

• Approve to send the change on as a pull request, and to deploy it if your project is set up to.
• Request changes to record what is wrong and set the attempt aside.
• Reject to throw the change away and put the ticket back on the board.

That is the whole loop. Runs and the approval gate explains each step in detail.

#If something does not work

If a screen says an integration is not connected, or an agent cannot start, tell your administrator. Those are set up on their side, not yours.