New | Enhancements | Changes | Bug fixes
New
Local headless test authoring from the CLI
If your team builds through coding-agent workflows, the mabl CLI now includes agent authoring skills that you can invoke from Claude Code, Cursor, or GitHub Copilot to create tests in headless mode. For example, after building out a new feature, you can start several local authoring sessions to run in parallel against the app on your machine to cover the new changes. Learn more.
Also, you can now authenticate the mabl CLI inside the same agent session with mabl auth login --auto. Learn more.
Smart search
With smart search, you can describe what you're looking for in plain language and include any exact text you recall, such as labels or fragments of a title. mabl then decides what to match by meaning and what to match by keyword. Learn more.
Measure mabl's value with the cloud MCP
The mabl cloud MCP's new value report returns measured counts — auto-heals applied, recovery sessions, AI-analyzed failures, and tests authored with the agent — so you can show what your quality automation is delivering. Because it comes through the MCP, your AI client can incorporate those numbers into your reports for leadership. Learn more.
New mabl API endpoints for reporting and governance
We recently introduced two sets of mabl API endpoints to help with reporting and access control:
- Test activity - retrieve test run counts, execution seconds, and maximum concurrency at the workspace or account level to build ROI and reliability dashboards in tools like Tableau, Power BI, or Looker. Learn more.
- User, seat, and role management - provision and offboard members, change roles, reclaim inactive seats, and inventory your workspaces, so access governance and license right-sizing can run through your identity tooling instead of manual audits. Learn more.
Enhancements
Bulk label and metadata management with the mabl MCP
New mabl MCP tools let you edit test and plan metadata in bulk from a natural-language prompt: add, update, or remove labels, rename and re-describe, or enable and disable tests and plans across hundreds of entities at once. Prefer scripting? The CLI's tests edit-metadata and plans edit-metadata commands do the same. Learn more.
Handle dynamic data with less setup
Cut the manual work for dynamic scenarios with these two updates to the mabl Trainer:
- Variables from a page URL - capture all or part of a URL straight from the address bar and store it in a variable. Learn more.
- Loop over elements or array items - new loop options help reduce the amount of setup and custom steps required for automating complex looping scenarios. Learn more.
A more flexible agent planning experience
Planning tests with the mabl agent gets more hands-on:
- Revisit planning sessions - reopen an earlier planning conversation to keep refining a test or to spin up a similar one. Learn more.
- Edit test configuration in the chat - set the app, environment, DataTables, credentials, and more yourself instead of prompting the agent toward them. Learn more.
Improved Selenium import
Instead of translating actions one at a time, the updated import records your Selenium test in detail as it runs and hands that recording to the mabl agent, which authors a complete mabl test in the cloud. The new Selenium import requires far less refactoring of imported tests before they're ready to run. Learn more.
Richer failure diagnostics
Two updates help you see what failed and why, faster:
- Failure columns in your results tables - new Failure reason and Failed at columns make it easier to scan where and why runs failed right from the results tables, without opening each run. Learn more.
- AI failure analysis in Slack and Teams - failure-run notifications now carry mabl's root-cause synopsis inline, so your team can start triaging without switching to the app. Learn more.
Reuse downloaded files in upload steps
File upload steps now open a mabl file picker in place of your operating system's dialog. Upload from your computer, reuse a file your test downloaded earlier in the same run, or pick one already saved to your workspace. Learn more.
Safari and Firefox for mobile web emulation
Mobile web emulation now runs on Safari (WebKit) and Firefox alongside Chrome, so you can confirm your app's mobile experience holds up across browsers. Learn more.
Changes
Deprecation of the local mabl 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 opening mabl in multiple browser tabs could log out all tabs simultaneously before the configured session timeout.
- Fixed a bug where deleting credentials assigned to tests caused those tests to fail at runtime; mabl now surfaces the affected tests before deletion and clears the reference.
- Fixed an issue where deleted tests appeared in a plan as "[Browser] undefined" and caused the plan to display 0 credits.
Test authoring
- Fixed a bug where database query editor values containing variables were cleared after saving or running.
- Fixed an issue where an assertion step targeting an HTML attribute silently reverted to an Inner text assertion after saving.
- Fixed a bug where a flow's loop configuration was lost when the test was reopened in the mabl Trainer and re-saved.
- Fixed an issue where visual assertions on downloaded files used a stale filename instead of the current DataTable variable value on each iteration.
- Fixed a bug in the API Test Editor where API tests failed to run with a workspace context error.
- Fixed an issue where the mabl Mailbox showed the mailbox address in the To field for emails received via BCC.
- Fixed a bug where importing an API test with multiple requests sharing the same ID could produce an invalid test, and added guards to prevent duplicate request IDs on import.
Test execution
- Ensured that clicking Rerun on a failed test within a plan run reuses the plan's configured settings and credentials instead of triggering an ad hoc run.
- Fixed an issue where runtime recovery still ran in a plan where the setting was disabled.
- Fixed a bug where visual assertions on downloaded CSV files could incorrectly pass row-count criteria that did not match the file.
- Fixed a bug where the Match regex assertion option in API tests failed when the expected value was provided through variable syntax, even when the value matched.
Test output
- Fixed an issue where echo step variable values were not resolved in exported PDF reports.
- Fixed a bug where the By Deployment results view loaded inconsistently and ignored pagination changes.
- Fixed an issue where the test status table on the coverage page would show tests that did not match the selected environment filter.
- Fixed an issue where the application name could display incorrectly on the test result page for cloud runs.
- Fixed a bug where the plan run results summary omitted step and failure details for certain failed runs, including in the mabl API response.
Integrations
- Fixed an issue where MCP tool calls could trigger duplicate test runs or test creations.
- Fixed a bug where mabl MCP tools failed to load for users without a default workspace set.
- Fixed an issue where Slack push notifications displayed raw markup instead of rendered hyperlink text.
- Fixed an issue where ad-hoc URL test runs were omitted from BigQuery exports when no environment was set.
mabl CLI
- Fixed a bug where reusable JavaScript snippets failed with "No javascript found" during an agent debug session replay.
Subscribe to our monthly newsletter
Interested in getting monthly updates sent directly to your inbox? Click here to subscribe to the monthly product newsletter.