August 2026 – mabl help

New | Enhancements | Changes | Bug fixes

New

Edit tests without launching your app

Not every test change needs a live browser session. Use these two new editing surfaces to handle changes that don't necessitate opening your application or the mabl Desktop App:

Direct editing from your AI coding agent - your coding agent reads the test's current steps, applies the change you describe, and saves it. Best for straightforward changes where your coding agent already has the context it needs. Learn more.
Quick edit in the web app - update steps, insert steps that don't need the app, reorder, or disable them, all in the browser. Learn more.

Set guardrails across your workspaces

Account admins and workspace owners get three new ways to manage their team at scale:

A company-level activity feed - company-scoped API keys, SSO connections, and admin privileges sat outside your workspace activity feed. Account admins can now monitor and audit these changes from the account dashboard. Learn more.
Enforceable credit limits - a workspace's credit limit used to be a guideline. Account admins now have the option of enforcing these limits, which blocks new cloud runs once that workspace uses its allocated credit limit for the billing period. Designed to help teams running mabl at scale from hitting surprise overages. Learn more.
Required cloud credentials - cloud credentials are write-only, so nobody can read a saved password back. Workspace owners can now require them for every new credential, and mabl enforces the policy in its API, so the web app, the mabl CLI, and the public API all follow the same rule. Learn more.

Loop until a condition is met

Loops used to need a count: a fixed number, a variable, an array length, or the number of matching elements on the page. While loops let a flow repeat for as long as a condition holds instead, so you can click "Load more" until it disappears or drain a queue without a workaround to exit the loop early. Learn more.

Enhancements

More coverage for file upload scenarios

Testing file uploads to your web app? mabl now supports three new scenarios:

Choose a file with a variable - combine literal text with mabl variable syntax to resolve a file variable at runtime, so one step uploads a different file depending on a DataTable value, an API response, or data on the page. Learn more.
Upload several files in one step - when an input accepts multiple files, the Trainer's file selection modal builds a list, and each entry can come from a different source. Learn more.
Hand a downloaded file to another test - one test downloads a file and another test in the same plan uploads it, so an export-and-reimport journey can stay two short, focused tests. Learn more.

Automatic Xray result syncing

Publishing a plan's results to Xray used to take a manual click after every run. Now you can set plans to sync on their own. Traceability keeps up without anyone remembering to push the button. Learn more.

Wait until an element is not present

The new wait until not present step pauses the test until the element leaves the page. Use it for tests where spinners, overlays, and toast notifications have to clear before a test can move on. Learn more.

JavaScript fallback clicks are now optional

A new application setting lets you choose how you want mabl to handle click steps that can't reach their target: attempt with JavaScript or fail the step. Learn more.

Easier tracking of branch merges

Previously, when a branch merged to master, any tests or flows that had picked up newer versions on master would get rolled back to the earlier version number that was formerly associated with the branch. Now the merge saves a new master version labeled with the merge so you can more easily trace edits and where they came from. Learn more.

Descriptions for reusable flows

You can now add an optional description that appears in the flows table, on the flow details page, and in the Trainer's flow picker. Use this field to add additional context for your team and to help the mabl agent make better decisions when importing flows. Learn more.

Saved basic auth credentials in the Trainer

Sites behind HTTP basic auth used to stop every Trainer replay with the browser's own authentication prompt. Starting with mabl Desktop App version 2.67.2, the Trainer applies a saved basic auth credential when it opens the browser. Learn more.

Other UX improvements

Two updates to the interfaces you work in every day:

A refreshed workspace dashboard - the top of your dashboard is now four cards: tests, test runs, latest pass rate, and flaky tests. Hover over any card for a breakdown across browser, mobile, API, and performance tests, then click a row to land in that report with the filters already applied. Learn more.
Dark mode in the Trainer - the Trainer used to render in the light theme even when the rest of mabl was set to dark mode. On the latest version of the mabl Desktop App, every Trainer window follows your preference. Learn more.

Changes

Retirements

The following were retired in the month of August:

mabl local MCP server - mabl retired the local MCP server on August 25, 2026. The cloud MCP server covers everything the local one did, plus workspace management, conversational test planning, failure investigation, and automatic updates. Learn more, including how to migrate from local to cloud.
Agentic runtime recovery - as of August 3, 2026, runtime recovery is off in all workspaces. Your tests, plans, results, and auto-heal are unchanged. Learn more about why we made this decision.
Segment integration - mabl retired the Segment integration on August 6, 2026. No action is required. Stay tuned for updates on native, agentic solutions that help you understand your app's coverage. Learn more.

Bug fixes

General

Fixed an issue where mabl's default plans appeared in the Add tests to plan dropdown and tests could be added to them.
Fixed a bug where label search on the Tests page and workspace search froze the browser in large workspaces.
Fixed an issue where the database queries page didn't show row actions until another page had been visited first.
Fixed a bug where the Release notes link on the desktop app update banner opened a blank window.
Fixed an issue where deleting a test from a plan reset the plan's device setting from mobile to desktop.

Test authoring

Fixed issues with step groups, including renaming, saving metadata, and disabling parent step groups.
Fixed an issue that prevented changing the current parameter value in a reusable JavaScript snippet, where the value reverted on save.
Fixed a bug where a recorded file upload step lost its element target, causing Set file input steps to fail with "Element not found" on replay.
Fixed an issue where the Configure Find option to click elements that aren't actionable couldn't be unchecked.
Fixed a bug where copied Find Element steps referenced the original elements.
Fixed an issue that prevented asserting whether an element is present or not present rather than asserting on its attributes.
Fixed a bug where adding or editing a scroll step in the Android mobile Trainer stopped all device interaction.
Fixed a bug where the authoring agent silently skipped reusable flows containing a JavaScript snippet, even when instructed to use them.
Fixed an issue where an agent test creation task remained shown as queued.

Test execution

Fixed an issue where a click that worked in the mabl Trainer failed during cloud runs.
Fixed a bug where triggering a plan containing a data-driven test whose DataTable had no matching scenarios produced a plan run page with no information; the affected test now reports a clear result and the rest of the plan runs.
Fixed an issue where an AI assertion on a downloaded spreadsheet treated the first row of values as a header row.
Fixed a bug where a variable used in a dropdown selection step resolved to "Unknown Value" during cloud runs.
Fixed an issue that prevented the local Appium server from starting, which blocked local iOS and Android testing.
Fixed an issue where AI assertions on downloaded files failed with "LLM call timed out".
Resolved issues that caused plan runs to hang for hours instead of failing.
Fixed an issue where a loop index variable used in an element locator resolved to the first element on every iteration.
Fixed a bug where a file assertion completed without returning a verdict, so the result never landed.
Fixed an issue where a click passed in a cloud run but the handoff to the newly opened tab failed on later steps.
Fixed a bug where certain storefronts rendered blank in cloud and local runs because of an interrupted page transition.

Test output

Fixed an issue where run logs and step descriptions could show asterisks in place of ordinary characters.
Fixed a bug where the Link Crawler broken links report was not generated and its download button was missing.
Fixed an issue where the Metrics tab was blank on browser performance test runs.
Fixed a bug where stopped performance test runs were reported as failed.

Integrations

Fixed an issue where the plan notification Starting URL showed the API URL instead of the app URL when both were configured.
Fixed a bug where the get_mabl_flow_steps MCP tool silently truncated large flows and dropped steps.
Fixed an issue that prevented mabl's cloud MCP tools from loading in Claude Desktop.
Fixed a bug where the Jira project dropdown didn't list all projects in organizations with a large number of them.

mabl CLI

Fixed an issue where the mabl CLI resolved the wrong environment URL.
Fixed a bug where plans list and tests list table output wrapped unreadably.

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

New