Pick the project, describe the change in your own words, press Go, follow it on Tasks while an agent works, then read what it did, what it cost and the evidence in its report, and use the change in the app.
Three sizes of work side by side. A sketch that hands back a clickable mockup with no code changed, the same task built once you like it, and an epic that is analysed, designed and planned before it is split into tasks and checked against the original request.
Start a task Guided, answer the agent's questions when it stops to ask, find every answer in Q&A, then talk an idea through in a chat with an advisor that has read the code, and complete it.
Write a request that comes back as you pictured it. Attach a sketch, list what finished means as a checklist, say what to leave alone, set who it is for and how urgent it is, then see the report check every item and the result beside the sketch.
Ask for a change that touches sign-in, and see the agent build and test it, then hold it on the Activity feed for a person. Read what it touches, approve it, watch it ship, and find the approval kept with the task.
Keep a week of your team's tasks in order. Keep an idea as a draft and start it later, stop work that went the wrong way and reset it, search the list, narrow it to a period and group it, decide what runs next in the queue, look back as a table, share what shipped as release notes, and see what a piece of work cost.
Set up recurring work once and let it file itself. Create a scheduled job that keeps an app's dependencies up to date every morning, say when it runs and for which project, let it run unattended, try it at once with Run now, follow the task it files to the finish, and look back over its runs.
Put the right kind of agent on each kind of work. Hand a question to a researcher and read its report, give slow tests to a test optimizer and read what it changed, put a small change on a cheaper model and see what Inherit means, then compare who did what and what it cost.
Add a project agents can work in safely. Connect a repository and pick the machine its work runs on, set the branch agents deliver into and the checks every change must pass, say who works in it, and have the Onboarding engineer get the repository ready for agents.
Give your team enough capacity, and see that it is healthy. Start a cloud machine from your own AWS account and watch it come online, see what a machine needs before it runs work, add places to work so tasks run side by side instead of taking turns, and read a machine's health and metrics.
Make sure everyone sees what they should, and answer the auditor. Give a group access to one more project, see everything one person reaches and through which group, check whether they may do one thing and why, find who made a change and when, and find the record of what an agent did in its session.
Fertig gelesen? Dann geben Sie die erste Aufgabe ab.
Die Agenten arbeiten auf Ihren eigenen Maschinen und liefern fertige Arbeit zurück, jedes Stück mit Bericht. Registrieren Sie sich und wählen Sie den Plan, der zu Ihrem Betrieb passt.