After Marketplace deployment, complete the tenant activation wizard in the controller portal. Typical time is about 10 minutes. Until the tenant is ACTIVE, main product routes stay gated.
Prerequisites
- Successful Azure Marketplace installation
- Controller portal URL — from deployment outputs, or from the managed resource group install ID:
acme→
https://acme-miso-miso.azurewebsites.net - Global admin or designated activation operator account (Microsoft admin consent runs in this wizard — not as a separate Entra configuration after Marketplace deploy)
Where it happens
The wizard runs entirely in the controller web portal in your Azure deployment. It configures tenant consent, enterprise identity, capacity preset, access model, endpoints, and final platform validation.
Integration developers begin workstation setup after activation — see Developer workstation setup.
Steps
Work through the wizard in order:
- Tenant consent — Accept terms and confirm your organization owns this deployment.
- Identity setup — Follow Connect Microsoft Entra ID. The authorized Microsoft administrator sign-in selects the directory; you do not need a pre-known directory identifier. Confirm the intended employee directory, which can differ from the Azure hosting directory.
- Size selection — Choose a capacity preset aligned with expected load (see Platform sizing).
- Access model — Configure how users and applications reach the platform.
- Access endpoint — Confirm URLs clients and integrations will use.
- Validation and deployment — Run platform validation; wait until activation reports success. Typical time for the wizard is about 10 minutes.
When finished, tenant status becomes ACTIVE and the main portal unlocks.
In the access-model step, configure a supported text model and a separate embedding model:
- Text model:
gpt-4o-miniorgpt-5.6-luna - Embedding model:
text-embedding-3-small
For OpenAI, these are model IDs available to the configured API key. For Azure OpenAI, they must match the deployment
names in the selected Azure OpenAI resource exactly. A ChatGPT workspace subscription does not by itself grant OpenAI
API model access.
If validation or the first assistant request returns 429, 400, 404, or another provider/model error, verify
both configured names before retrying. After the names and provider access match, a 429 normally means provider quota,
rate-limit, or regional-capacity exhaustion; check the provider metrics and quota rather than changing the channel.
If you are logged out, or Entra and local admin sign-in both fail, wait about 20 minutes and try again. Services continue to install in the background. That work can change hostnames and entry URLs. Wait until all services are up before signing in again — do not treat a failed login in that window as a broken account.
What done looks like
| Wizard step | Business outcome |
|---|---|
| Consent | Tenant claimed by your organization |
| Identity | Enterprise sign-in path ready |
| Size | Capacity preset selected |
| Access model | Who may authenticate and how |
| Endpoint | Stable URLs for users and integrations |
| Validation | Platform services healthy (~10 minutes typical) |
If activation fails, note the step and error text — fix prerequisites (identity registration, DNS) and retry from that step.
Validate
- Portal shows tenant status ACTIVE
- Admin can open main portal routes without onboarding redirect
- OpenAI or Azure OpenAI validation uses
gpt-4o-miniorgpt-5.6-lunafor text and
text-embedding-3-smallfor embeddings; Azure deployment names match exactly - Optional: First session checklist
Limits
This sequence describes the Azure customer-owned deployment. The portal receives its steps from the installation mode; other modes can show a shorter sequence. Follow the displayed steps rather than looking for absent Azure capacity controls. Ten minutes is typical for the wizard UI; background service install can continue after that, and sign-in may fail until those services are up (about 20 minutes).
Common mistakes
| Mistake | Fix |
|---|---|
| Closing wizard before validation | Complete the final validation step |
| Wrong identity permissions | Fix identity inputs; re-run validation |
| 429 or provider/model error | Verify the text and embedding model IDs first. For Azure OpenAI, verify exact deployment names. If they match, check quota, rate limits, and regional capacity |
| Using a ChatGPT subscription as OpenAI API access | Configure an OpenAI API key with access to the selected models, or configure an Azure OpenAI resource and matching deployments |
| Starting integration work before ACTIVE | Finish wizard; developers use Build path after activation |
| Retrying login immediately after logout | Wait about 20 minutes; backend install may still be changing URLs |