Test your Copilot Managed Runtime partner integration end to end

Use this checklist to test your Copilot Managed Runtime partner integration end to end. Run it for the first successful partner integration, and repeat it after CLI, gateway, or orchestration changes. Use a disposable app and keep one source mode from initialization through iteration. Replace placeholders and run app-scoped commands from the app root.

Stage Action Observable pass condition
1. Readiness and authentication Confirm the partner Microsoft Entra application and embedding domain are allowlisted, the required packages are installed, and the app maker is signed in through the partner. In hosted tests, send a tenant-scoped read such as ms app list --json through HTTP Forward Mode. ms --version succeeds; allowlisting is recorded for the partner application and applicable play-host origin; the CLI read returns JSON for the intended user and tenant without running ms auth login or persisting a CLI token cache.
2. Initialize the app Run the documented ms app create or ms app init branch for platform-managed Git, external Git, or --repo none. Persist the generated configuration. ms.config.json exists and records the selected source mode, and ms app info --json resolves the new app's server-side state.
3. Build the generated project Install dependencies and run the project's configured build command. The command exits successfully from a clean checkout and produces no compile or SDK integration errors.
4. Add governed data Run ms connector list --json, select one connector permitted by current tenant policy, and follow the documented add data-source flow. Persist the resulting ms.config.json, generated/, and .ms/schemas/ changes. The add command succeeds, the generated files survive a clean checkout, and the selected operation returns expected test data for the signed-in maker.
5. Run Local Play Start ms app dev (or the documented config-only variant) and embed the exact Local Play URL emitted by the CLI in the partner experience. The Copilot Managed Runtime app host and nested app render from that emitted URL, and the connector operation works; the partner does not construct or fall back to a bare dev-server URL.
6. Submit and monitor a hosted build Git-backed branch: push the full commit, then run ms app build --app <appName> --commit <fullCommitSha> --no-wait --json. Retain returned identifiers and poll ms app build-status --app <appName> --commit <fullCommitSha> --show-log --json. Artifact branch: no hosted build applies; continue with the successful local build. Git-backed submission returns immediately, status reaches terminal success, and progress or logs remain visible. Artifact mode records this stage as not applicable without claiming a hosted build.
7. Preview without changing live Git-backed branch: run ms app play --mode preview --commit <fullCommitSha> --no-browser --json and open the returned preview URL. Artifact branch: use the successful Local Play result; the documented artifact flow has no Git-backed hosted-build preview step. The candidate version marker appears in preview while the live URL still shows the prior deployment, or remains undeployed on the first run.
8. Deploy explicitly Git-backed branch: run ms app deploy --commit <fullCommitSha> --json. Artifact branch: follow partner artifact deployment: verify the local build with ms app pack, then use ms app deploy --json to pack and upload it, or ms app deploy --artifact <artifactPath> --json for a prebuilt ZIP. Deployment reports success only after an explicit deploy action; the artifact branch has the required administrator setting enabled and uploads the intended local build or prebuilt ZIP.
9. Share and open live Run ms app share <principals> --access play, then obtain the live URL with ms app play --no-browser --json and open it as a shared test user. The intended principal can open the deployed version at the returned stable runtime URL; an unshared control user cannot.
10. Prove one iteration Change a visible version marker and repeat the applicable build, preview, deploy, share, and open checks without creating a new app. The runtime URL is unchanged, now serves the new deployed marker, and no longer serves the prior version as live.

Failure evidence for triage

Capture:

  • The exact command and UTC timestamp
  • Logical app name and environment
  • Operation, build, request, and correlation IDs returned by the CLI or gateway
  • Sanitized JSON output, status, and relevant logs or error text
  • The failing stage and source mode

Never capture or share access tokens, hop tokens, passwords, connection secrets, or secret-bearing headers.