Everything, in reading order
Filter by where you are. Every row opens a page that exists — there are no placeholders in this list.
Five ways in: run a business task, work a website with no API, connect an inbox, automate a recurring process, or build software.
The jobs people hand over first, and what each one needs from you before it can run.
Start from a process that already works, then describe the way yours differs.
What connects over an API, what runs in a browser instead, and why the difference matters.
Supplier portals, booking systems, back-office tools: how Kaer works a website in its own browser, where you sign in, and where it stops for your approval.
Find what's overdue, draft each reminder in your tone, and approve it before it goes. A step-by-step guide to chasing invoices with Kaer.
Set up Kaer Mail, let it sort the morning's enquiries, approve drafted replies before they send, then choose how much runs automatically.
Attach the spreadsheet, describe the tool, check the live preview and publish it on its own address. A step-by-step guide to building tools with Kaer.
Describe the report once, run it by hand, add an approval step and put it on a schedule. A step-by-step guide to recurring work with Kaer Workflows.
Briefing a run, reading a diff in two minutes, and where to put approval gates.
Describe an outcome. It plans, works, and stops at the step that carries a consequence.
Company processes that keep running after you close the tab, on a schedule or a trigger.
Triage, replies drafted in your tone and held, and the point where an email becomes an automation.
An agent with its own machine, for the software that has no API — and what contains it.
The internal tool you never got round to building, with a development team or without one.
A role, a set of access, and the rules it works under — briefed once, delegated to from then on.
One workspace, several people, and four different kinds of approval.
Authentication, the REST surface, error shapes and rate limits.
One endpoint, 4 lanes, from $0.50 per 1,000 queries on a prepaid balance.
What each action costs, and the meter that stops before the bill surprises you.
How a mistake is kept cheap, what a browser session can and cannot reach, and what is never training material.
What shipped, dated, including the parts that broke.
Every provider's price per thousand queries, and the surcharges that quietly move it.
Why the cheaper option stops being cheaper the moment an agent has to open a page.
Point Claude, ChatGPT, Codex or Cursor at Kaer's MCP server and hand it your search index.
The field guide is the one that changes how you brief it.
Six chapters on the difference between telling an agent what to do and telling it what you want back: framing the outcome, letting it plan before it touches a file, reading the diff, and putting gates where a mistake would actually cost something.
