Agent setups in Verging Memory CI

An agent setup is the harness, model, and agent-facing version of your product that Verging Labs tests. It includes the tools, tool descriptions, parameter schemas, and agent instructions presented during testing.

Verging Labs records your agent setups during onboarding. Your onboarding page gives you the exact setup names to use in release requests.

List your agent setups

Use the API to see the setups already recorded on your account:

curl -H "Authorization: Bearer $VERGING_API_KEY" \
  "$VERGING_API_BASE/v1/environments"

The relevant fields are:

{
  "environments": [
    {
      "name": "Claude Code Opus 5",
      "harness": "claude-code",
      "model": "claude-opus-5",
      "activated_suites": [
        "onboarding",
        "core-recall",
        "preference-adherence"
      ]
    }
  ]
}

Copy the name value exactly when submitting a release. If no setups are listed, or if the harness, model, tools, schemas, or agent instructions have changed, contact your Verging Labs onboarding contact before submitting. Verging Labs will record the setup you should use.

Activation

Each test suite is set up once on each agent setup before that combination can be used in a release. The activated_suites array lists the suite values already set up on each setup. Verging Labs sets up additional suites on request.

A release that names a suite not yet activated on one of its agent setups is declined with a clear message and nothing is used or billed. Submit it again once Verging Labs tells you the suites are set up, or name only the suites already set up (GET /v1/environments). A release is never held for activation.

Test several agent setups together

Send one or more setup names in the API's environments array or the GitHub Action's comma-separated environments input. One release produces one report with a summary and a section for each setup. Its release verdict reflects the least-ready setup.

Every explicitly selected suite must be set up on every named setup. When suites is omitted, Verging Labs selects the suites chosen for your account that every named setup supports and has set up. See API setup for mixed-setup behavior.

Environments tested

Regression reports show Environments tested, which counts each test suite exercised in each agent setup once. Plan accounting separately uses weighted Environment usage. See Billing for the calculation.

See Integrate Verging Memory CI with GitHub Actions, API setup, and Read a regression report.