Your first browser session.

01 / Get set up

You need Chrome or Firefox installed, a paid gripui licence, and an AI agent if you want it to drive for you. You can also use gripui directly in a terminal.

Follow the installation guide, activate your key, then install the agent skill:

gripui teach

Start a new agent session after installing the skill. Ask it to use gripui and choose a dedicated profile for this work.

02 / Choose your browser profile

In your terminal, attach a named profile in the project directory you want to work from. This creates or selects a separate browser context and records the project’s choice.

gripui attach first-session

Keep the same terminal session for these commands. There is no implicit default profile. If another job owns the profile, use a different name or have that job detach.

03 / Look at a page

Start with a public page that needs no login. Run each command separately and read its output:

gripui go https://example.com

gripui hud

The HUD lists the available actions with numbers. Those numbers identify controls in this tab and frame. They are observations from the page, not permanent IDs you should memorize.

04 / Make one move

Choose the link’s number from your own HUD, then run gripui do <number>. The example below illustrates the command shape; replace 3 with the number you actually see.

gripui do 3

Read the new page with gripui hud. This is the loop: observe, choose an action, then check what changed. A changed action set can produce a nonzero exit while still showing usable current refs; read the explanation.

05 / Give your agent a job

Ask your agent to use the gripui skill to read a public page and summarize what it finds. Be specific about the site, the desired result, and when it should ask you for help.

When you’re done, gripui detach releases your job’s claim without closing the browser. gripui quit closes that profile’s browser session.