Manual runs
Regardless of if you have automatic runs enabled, you can always manually run tests. This is useful when initially building out a test to ensure it’s running correctly.
To manually run a test, go to Tests → choose a test → Rerun. You can also choose to subscribe and be alerted when the rerun finishes. This requires having Slack notifications enabled.

You can also choose to rerun a test on a previous build by choosing Rerun → Rerun on build.

To rerun all failed tests for a previous build, go to the home page, hover over the latest build, select View build history, and choose the build.

Running several tests at once
Section titled “Running several tests at once”Use Run All to run a group that doesn’t run automatically, such as all tests
with the regression tag before a release. On the Tests page it runs
against the latest build. To run against an earlier build, open it from View
build history and use Run All there. Pick a
tag to run only those tests, or leave
it empty to run everything.

You can also skip tests that have already run on the build, which fills in the gaps without repeating anything. A test counts as already run if it passed, failed or errored there, with any version of its spec. Cancelled runs don’t count.
If you aren’t an admin and a channel runs tests automatically, Run All on the latest build only runs tests that haven’t run on it yet.
The MCP server supports the same choices:
test_runs_trigger takes a tag and skipIfRunOnBuild.