Glossary · Execution

Co-drive

Clicking and typing on a run’s live browser yourself, with coordinates expressed as fractions of the viewport.

What is co-drive?

Co-drive is direct interaction with a session the agent is holding: a click, a fill, a key, a scroll or a navigation, sent as an event whose coordinates are fractions of the viewport rather than pixels, so they are hit-tested server-side against the observed element map. There is no "done" call — the run re-checks the page after each interaction and resumes by itself once the block clears.

Why it matters

Some steps are neither automatable nor reducible to a code — an unusual consent dialog, a one-off confirmation. Co-drive is the escape hatch that keeps those from being a dead end, and fractional coordinates are what make it work from a screenshot of any size.

Related terms

See it in context

Back to the full glossary

Put your agent to work. Keep the decision.

Give your agent a real browser, authorize the accounts it may use, and keep control of every step. Repeated runs compile into skills that replay without a model.