Getting Started

The quickest way to get useful work out of Kaer without learning every page at once.

first-turn

<Chat prompt="Fix the flaky auth test and open a PR" />
→ operator plans 3 branches
<Branch id="b1" task="reproduce failure" />
<Branch id="b2" task="patch retry window" />
<Branch id="b3" task="update the test" />
→ branch b2 ships: PR #214 ready for review

The cleanest first run is simple. Sign up, let the onboarding wizard set the workspace up around you, run one turn in Chat, and review the result. Kaer becomes much easier to understand once you have seen one task move from request to outcome.

Create an account at kaer.ai. The free Starter tier needs no card and includes daily operator credits, so your first sessions cost nothing. On first sign-in, a short onboarding wizard asks what you do and what you want Kaer for. Answer it properly — it personalises the workspace around your kind of work, so the defaults you land on actually fit.

Your first turn

turn-1

<Chat prompt="Rename the billing
module and fix imports" />

Describe the outcome

One plain-English message in — the operator plans and does the work.

turn-2

→ 14 files changed
→ preview ready
<Review action="approve" />

Review it in place

The result comes back as something you can inspect, edit, and approve.

Your first turn happens in Chat. A turn is one message in, one outcome out: you describe what you want, the operator plans it, does the work, and returns something you can review in place. You can watch it think, interrupt it, and steer mid-task — see The Operator for how to read it while it runs.

A good first prompt names the outcome, the audience, and the tone. “Draft a launch summary for a B2B analytics release and keep it calm and concise” is far easier to steer than a vague request for “marketing help.” If the result is strong, stay in the same thread long enough to refine it once rather than starting again elsewhere.

The dashboard is the hub

Between turns, the dashboard tells you what is already in motion. If the workspace is shared, you can see at a glance whether tasks are active, which services are connected, and how much credit remains on the current plan. Work starts in Chat; the dashboard is where you keep an eye on all of it.

What to ignore at the start

You do not need to learn Assistants, Automations, Workflows, Macros, Mail AI, and Connectors on day one. Those pages make more sense after you have seen what a strong single-task session looks like. Most teams get better early results by learning the manual flow first and only turning it into repeatable infrastructure later.

What a good first result looks like

A strong first session gives you something concrete enough to review, edit, or send onward. It might be a summary, a report draft, a piece of copy, a research note, or an automation outline. If the result already feels close to finished, you are using Kaer correctly. If it feels directionless, narrow the request and keep the scope specific.

Where to go next

Once you have one result you trust, read The Operator to get fluent in turns, modes, and approvals, and Agent Workspace for how a single thread of work holds together. If the work needs a reusable specialist role, continue into Assistants. If it should stay grouped by client, project, or operating stream, go to Workspaces. If you already know something should run on a schedule, jump straight to Automations or Workflows.