Skip to main content

July 2026

Molly Exten
  • Updated

New | Enhancements | Changes | Bug fixes

New

Edit existing tests with the mabl agent

Now you can describe a test edit in plain language and let the mabl agent take care of it. When you edit with mabl, the agent agent opens the test, applies your changes, runs it to confirm it still passes, and reworks anything that fails before saving. Editing sessions run in the backgroud, so you can hand off a batch of edits and keep working: a new assertion, a stale selector, or a test that broke when your app changed. Learn more.

Agentic test authoring also picked up several other improvements this month:

Two recent updates make it easier for your team to track failures in your app:

  • Custom failure reasons - create up to 50 failure reasons that match how your team actually talks about failures, so known and expected failures stay separate from actionable regressions instead of skewing your pass rate. Learn more.
  • Automatic failure categorization - set up your workspace to automatically assign failure reasons for you. Learn more.

Manage users across every workspace

Reviewing who has access to what used to mean opening each workspace one at a time or scripting against the mabl API. Now account admins can see every user in the account from one dashboard, change user roles, remove access, and manage account admins. Learn more.

Teardowns for step groups and flows

Tests that create data can leave it stranded in your environment when a run fails partway through. Now you can mark a step group or flow at the end of a test as a teardown to clean up that data. An alternative to "always run" plan stages, teardown steps run at the end of the test, regardless of whether the test passed or failed. Learn more.

Enhancements

More ways to reuse files and their contents

Three updates make file-heavy tests easier to build and to maintain:

  • File variables - upload a file once and reference it from any upload step in the test. Learn more.
  • File contents as a variable - capture the text inside a downloaded or uploaded file as a string variable, then use the artifacts your app actually produces to drive later UI steps, API requests, or JavaScript snippets. Learn more.
  • Broader upload recording in the Trainer - the Trainer now captures a file upload step no matter how your app opens its file picker, including hidden inputs, inputs created on the fly, custom upload buttons, and inputs inside an iframe. Learn more.

Slack and Teams notifications for agent sessions and Xray syncs

Keep your team on track with new notification types in Slack and Teams:

  • Agent session notifications - choose which outcomes you want to hear about: ready for review, needs your input, or failed. Learn more.
  • Xray sync failures - when a plan's results fail to sync, the message carries the exact error from Xray and a Retry sync button, so nobody has to check Xray for a missing Test Execution. Learn more.

More flexible, more resilient custom find steps

The next time you create a custom find step in the Trainer, you'll notice some new options:

  • Playwright locators - a single Locator field accepts CSS, XPath, and Playwright locator syntax in place of the separate selector-type and ordinal controls. If your team already tests with Playwright, you can reuse those selectors as is. Learn more.
  • Optional visual find backup - configure custom find steps to fall back to a visual description of the element when the app layout shifts and the locator can no longer pinpoint the element. Learn more.

Two updates extend what mabl Link can reach and show you what it's doing:

  • Monitor your agents and tunnels - the recently updated networking page in the mabl app reports data on active agents, active tunnels, throughput, and more, giving you more visibility into your workspace's mabl Link activity. Learn more.
  • Reach localhost services - point a cloud browser test at localhost or 127.0.0.1 and the request routes through the Link tunnel to the machine running your Link Agent, so you can test a local development build without deploying it or exposing it to the internet. Learn more.

Actionability checks for click steps

By default, mabl clicks an element even when it's disabled, hidden, covered by something else, or still moving on the page. That behavior is convenient, but it can hide a real bug when a test passes on a button no user could reach. Turn on actionability checks and a click step fails unless its target is enabled, visible, unobstructed, and stable. Learn more.

Fast play in the Trainer

Replaying a long test from the beginning to reach a later step can take time. Fast play moves through steps quickly, so you can confirm everything still works or get your app into the right state sooner. Learn more.

See your deployment results at a glance

The updated output page for a deployment event includes more details on the results, reruns, and the final status. Learn more.

Changes

Retirement of runtime recovery

As of August 3, 2026, runtime recovery is off in all workspaces. Based on the outcomes from the early access program, we've decided to take a different approach to this issue so you can get failed runs back to passing faster, with less time spent sorting through results. Learn more.

Deprecation of the local MCP server

To focus continued investment on the cloud MCP server, mabl deprecated the local MCP server on May 18, 2026. The local server will continue to function, but it will no longer receive new tools, feature updates, or support. Learn more.

Bug fixes

General

  • Fixed an issue where the application filter did not apply on the Tests page.
  • Fixed a bug where a test that existed only on a deleted branch remained in the workspace but could not be opened or edited.
  • Fixed issues with loading the test details page.
  • Fixed a bug where the plan-level Run and Stop controls reflected only the most recently completed stage instead of the whole plan run.
  • Fixed an issue where the mabl Desktop App update banner rendered as a blank bar instead of showing the update link.
  • Resolved issues with launching the Desktop App on certain Linux and macOS systems.
  • Fixed an issue where the activity log did not render correctly while a Link Agent was active.

Test authoring

  • Fixed an issue where Chrome's Save password prompt appeared during training and local runs.
  • Fixed a bug where Configure Find was unavailable on newly created native mobile assertion steps.
  • Fixed a bug where assertion steps failed to save in the mabl Trainer after upgrading the Desktop App to 2.57.0.
  • Fixed an issue where the mabl Trainer's empty-flow save error didn't identify which flows were empty.

Test execution

  • Fixed a bug where a switch context step could not find a newly opened tab on replay.
  • Fixed an issue where an API step assertion using a JSON path with a dot before square brackets failed in cloud runs.
  • Resolved performance issues with large DataTables.
  • Fixed an issue where the a custom find step rejects a valid locator match, falls back to visual find, and clicks on the wrong element.
  • Fixed issues that caused cloud runs routed through a Link Agent to time out or hang.

Test output

  • Fixed an issue where a manually recategorized failure reason still showed the stale auto-generated category on the results page.
  • Ensured that failures categorized as Performance appear on the coverage page.

Integrations

  • Fixed a bug where an Xray plan run sync could remain stuck in a pending state.
  • Fixed an issue where the get_mabl_tests MCP tool returned only browser tests, omitting API and mobile tests.
  • Fixed a bug where certain mabl MCP operations returned 403 errors when the server was authenticated with an Editor-level API key.

Subscribe to our monthly newsletter

Interested in getting monthly updates sent directly to your inbox? Click here to subscribe to the monthly product newsletter.