Release notes — September 2026 (29–30 September)

Summary

Performance testing is now sold as a paid add-on to a Professional, Custom or Enterprise licence, in four packages that set the load limits. A Trial still includes it. Test-data specialists can fill a project's existing tests from a data set in one pass, with a preview first and Undo afterwards, from the Data Sets screen or by asking the assistant. Load runs gain a Server log tab, and authentication now reliably reaches every load request. See API load testing and the add-on packages.

Breaking changes

Performance testing is a paid add-on

Performance testing is a paid add-on to a Professional, Custom or Enterprise licence. No edition includes it on its own, Enterprise included. This replaces the edition limits announced in the 21–28 September release notes. The add-on's package sets the limits:

• Trial includes performance testing with the Scale limits, so it can be evaluated.
• Citizen Developer (Free) sees a preview of the Performance screens and cannot buy the add-on.
• The limit is per load generator. A run split across load agents can send more in total.
• Where to see what you have. The Performance home shows your package and its limits. Settings → License lists it under Add-ons, with its end date if it was bought for a shorter term than the licence.
• After buying or renewing, the product picks up the change at its next licence check, within 4 hours. An Administrator can press Refresh on the License page to check at once.
• When the add-on ends, scenarios, runs, reports and targets stay readable. New runs, scheduled runs and changes to scenarios are refused with a message that says when the add-on ended. Administrators see a warning on the Performance home from 14 days before the end date.

What to do: to add performance testing to your licence, contact sales. Add-on pricing is quoted per customer.

Features

Fill tests from a data set

• Push test data into existing tests. Choose Fill tests… on the Data Sets screen (or Project Settings → Data Sets) to send the values of a CSV, Excel, JSON, YAML or XML data set, or a typed table, into every existing test of a project, one test at a time. See Fill tests from a data set.
• Two modes. Fill values writes one row into each test, sharing rows out across an endpoint's tests (or giving every test the first matching row). Run each passing test once per row links the tests to the data set for data-driven runs.
• Only fields a request already sends. Columns are matched to fields by name, dotted columns fill nested body fields, and values are converted to each field's type. Path, query, form, JSON body, GraphQL variables, JSON-RPC and WebSocket-RPC params, SOAP parameters and ordinary headers can be filled. Unused columns are listed.
• Deliberate values are kept. A boundary or wrong-type value only one test sends, {{linked}} values, credentials and transport headers are left alone. INVALID_ columns never reach a test that must pass, and an EXPECTED_STATUS column decides which rows a test may take.
• Checks move with the data. A check that repeats a value the request sent is updated to the new value.
• Preview and Undo. Every change is previewed before anything is written. Recent fills lists earlier fills with Undo, which restores every test the fill changed except one edited afterwards, and names it. A filled value counts as your edit, so regenerating tests keeps it.
• From the assistant. Ask the Project Assistant to "fill our tests with the Users data set" or "undo the last fill". It shows the same preview as an offer, and nothing is written until you accept.

Improvements

Performance testing

• A server log for every run. The run report has a Server log tab listing what the server did for that run, in order: the plan, the check before load, each stage, the first time each kind of failure appeared, and the end. It can be filtered to warnings and problems. The log is not part of exports, share links, the assistant or the MCP server.
• Credentials reach every load request. Authentication from your profiles is applied to each load request, and tokens renewed mid-run keep each step's overrides. If a profile resolves but its credentials do not reach a request, the run warns you instead of producing a report full of 401 responses.

Fixes

• Test generation: the progress of Generate Tests for All is shown again when you leave the screen and come back; generation kept running before, but the progress bar was lost.
• Test editor: a test reopened right after saving now shows the saved values the first time, not on the second reopen.

Known issues

• WebSocket-RPC calls and MCP server endpoints cannot be load-tested; each needs a held session per user.
• Performance (load) tests do not use an environment's trusted CA certificate setting yet.

Related links

• Performance & Load Testing add-on packages
• API load testing overview
• Performance testing documentation
• Fill tests from a data set
• Release notes — 21–28 September 2026