Product · MCP
One line in your agent.A program in your account.
Connect Claude Code, Codex or Cursor to mcp.kaab.ist. Your agent opens a real Kaabist browser, works through the site with you, and what it does becomes a program: versioned, linted, and tested with a dry run before it ever submits anything.
- 01You ask your agentIn Claude Code, Codex or Cursor, you describe what you need from a site, in plain words.
- 02It opens a real browserOver MCP, the agent opens a Kaabist browser and reads the page as a view with ranked locators, not as pixels.
- 03It acts on the pageIt fills, clicks and waits the way you would, and clears whatever covers the page first.
- 04It tries the extractionFields are tried on the live page, so the agent sees real values before anything is saved.
- 05What it did becomes a programThe session turns into a versioned program: the steps, the locators with their fallbacks, the typed fields.
- 06Tested, then publishedIt is linted and tested as a dry run, then pushed. From now on it runs without the agent, on a schedule or a call.
How it works
01
Connect once
The server speaks Streamable HTTP and signs your agent in with OAuth on your Kaabist account. You decide what it may do. It starts in test mode, on drafts, with a credit cap.
exampleclaude mcp add --transport http kaab https://mcp.kaab.ist/mcp
02
A real browser, not a guess
The agent drives a browser on Kaabist's own pool and reads a condensed view of the page: forms, lists, landmarks, and ranked locator candidates for every element. You can watch it work in the panel, or take over at any moment.
example› browser_open https://portal.example.test/login › browser_page_view forms: login (#email, #password, button "Sign in") › browser_fill e12 "{{inputs.email}}" › browser_click e1503
From session to program
browser_to_program turns what the agent did into a draft program. lint_program checks it. test_program runs it as a dry run that stops before the first step that submits anything. You approve before it is published.
example› browser_to_program → draft prg_4Rb v1 (7 steps) › lint_program → 0 errors, 1 warning › test_program → dry run stopped before click Renew
Questions
Askedbefore you ask.
More of Kaabist
Resumes where it stopped
Some steps only you can finish: a one-time code, a captcha, an approval. Kaabist pauses the run right there and emails a link to you, or to the teammate or customer you choose. Once they answer, the program continues from its last checkpoint. Nothing starts over.
Never submits twice
Mark the steps that change something (a submit, a payment, an order) as side effects. Kaabist writes each one down just before it runs and again once it has worked. If a server crashes in between, Kaabist never presses the button again on its own: it pauses and asks you to check.
Self-healing locators
Every target is recorded with several ranked locators and a fingerprint of the element it found. When the site changes and they stop matching, Kaabist finds the element again, writes the fix down as a proposal, and lets you promote it to a new version.