Kobiton 4.26.4S Release Notes

Release date to be confirmed

New features and major changes

Device-only view for live sessions

View just the device screen during a live session, without the surrounding Kobiton interface. Device-only view hides the sidebars, headers, and panels and sizes the viewport to the device screen. It works for phones and tablets, on Android and iOS, in portrait and landscape.

Device-only view is useful for demos, screen sharing, and observing automation runs.

Highlights

  • Activate device-only view by adding &view=device-only to a live session URL.

  • In device-only view, only touches and swipes are sent to the device. Keyboard input is not sent to the device under test.

  • A floating exit button ends the session and releases the device. When an Appium session is running, the button instead closes the live view and keeps the automation running.

  • For AI-assisted workflows, you can retrieve the live view and the device-only view from the agent during a live session.

Availability

Available for in-progress live sessions.

Guide AI-driven sessions from the device canvas

During an AI-driven session in device-only view, tap or swipe on the device canvas to steer the device. The agent observes your touch and swipe input between its scripted commands and acknowledges it.

Highlights

  • Tap or swipe on the device canvas in device-only view to signal intent to the AI agent.

  • A new getUserInputEvents MCP tool lets the agent read these canvas events.

  • Only canvas touch and swipe are sent to the device. Keyboard input and other interactions remain disabled in device-only view.

Availability

Requires device-only view.

Bug fixes and improvements

Appium and automation

  • Fixed an issue where recorded text-entry steps were omitted from generated Python (pytest) scripts, causing form submissions to run against empty fields.

  • Fixed an issue where generated Python (pytest) scripts could fail at the first screen transition after a form submission.

ADB shell

  • Fixed an issue where the am start command was rejected as an unsupported command when launching protected system apps such as Chrome and Settings.

App management

  • Fixed an issue where installing an app version whose file had been removed from storage returned an unclear error. Affected versions now return a clear, actionable message.

Device management

  • Added the OS build number to the device details modal for Android and iOS devices.

  • Added an OS build number filter to Device List search.

Manual sessions

  • Fixed an issue where command latency was missing or calculated incorrectly for sessions that reconnected multiple times.

MCP server

  • Added a getAppParsingStatus tool so AI agents can confirm an uploaded app has finished processing before they use it.

  • Fixed the listSessions tool ignoring the limit and offset parameters. Agents can now control how many sessions are returned.

  • Removed misleading app expiry fields (expiry_date and is_expired) from listApps and getApp responses. Uploaded apps remain usable and do not expire.

API and CLI

  • Improved the error message returned when app signing cannot proceed, replacing a generic server error with a descriptive message.

SSO sign-in

  • Added a clear error message on the SSO sign-in page when an organization’s SSO name is invalid.