Guides

A guide walks through one task from start to finish, with code that Kaabist's CI runs on every change. Pick the task you have; each guide says when it is the right approach and when another one is.

Signing in and asking people

• Authenticated flows: a login with a one-time code, with secret inputs, await_input, a saved session, and a second run that skips the login.
• Hosted resolve pages: let your own end users type a code or unblock a parked job on a page that needs no account.

Reading data

• Lists and pagination: lists of objects, "next" links, infinite scroll and the lint rules that keep them bounded.
• JSON and network locators: read the JSON a page embeds or fetches instead of the HTML drawn from it, and move to the http tier.

Research

• Research: ask a question, name the fields you want and where to look, and get values copied verbatim from the pages with their sources.
• Search: ranked web results through Kaabist's own engines, held to the domains you allow.
• Deep research reports: an open question answered in sentences that each cite the passages they rest on.

Running on your behalf

• Schedules and change detection: watch a page on a cron and hear only when a watched field changes.
• Webhooks: receive Kaabist's events, verify their signature and process each one once.
• Bring your own providers: use your own proxies, language models and captcha solvers.

Safety and quality

• Testing: test keys, cost estimates, lint and replaying a version against a stored page.
• Side-effect safety: programs that submit forms or place orders, and how Kaabist avoids doing it twice.
• Security and data: what Kaabist stores, for how long, and what it never keeps.
• Drift and canaries: how Kaabist notices a program quietly returning worse data before your next job does.
• Migrating: moving existing scrapers and browser scripts to Kaabist programs.

Tools

• SDKs: the TypeScript, Python and Go clients.
• CLI: the kaab command line.
• MCP: connect a coding agent to Kaabist.