Connect Your First AI Provider: A Setup Checklist for DAVE

Provider configuration is restricted to the Admin role. No other role, including Create, Curate, Use, Reporting, or Financial, can view, create, edit, or delete providers. If you do not have the Admin role, ask your tenant Admin to complete this setup or to elevate your role.

Before opening the provider form, collect the following from your AI service account:

• Your API key. This is the secret credential issued by the AI service (for example, an OpenAI secret key starting with sk-, an Anthropic API key, or a Google AI Studio key). Keep it in your password manager until you paste it into the form: you will not see it in full again after saving.
• Your base URI, if required. For OpenAI, Anthropic, and Google AI, DAVE pre-fills the correct default endpoint. For Azure OpenAI, OpenAI-Compatible services, and Custom providers, you must supply your own endpoint URL. Locate this in your service's dashboard before starting.
• The model or models you intend to use. You do not need to know the exact model IDs in advance if you plan to use Fetch from API mode, but knowing which models your team will use helps you curate the catalog quickly after fetching.

DAVE supports seven provider types: OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, Local (Ollama), and Custom. The checklist below applies to all of them. Provider-type-specific notes are called out where the steps differ.

Work through each item in order. Check off each item as you complete it.

Access and navigation

• Confirm you hold the Admin role in your DAVE tenant. If you are unsure, navigate to Admin, then API Management, then Providers: if the page is inaccessible, you do not have the Admin role.
• Navigate to Admin, then API Management, then Providers. The page shows a card for each existing provider, or an empty state if none have been configured yet.
• Click Add Provider in the top-right corner. The create form appears above the provider list.

Name the provider

• Enter a descriptive Name that identifies the service, environment, and purpose. Examples: "OpenAI Production", "Anthropic Dev", "Azure OpenAI eastus2 gpt-4o". Avoid generic names like "Provider 1": the name appears in every agent configuration dropdown and must be immediately recognisable to any team member.
• For Azure OpenAI providers, include the deployment region or deployment name in the Name field. Azure endpoints are deployment-specific and two Azure providers pointing at different deployments are otherwise indistinguishable by name alone.

Select the provider type

• Open the Provider Type dropdown and select the type that matches your AI service: OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, Local (Ollama), or Custom.
• Confirm the Base URI field after selecting the type. For OpenAI (https://api.openai.com/v1), Anthropic (https://api.anthropic.com/v1), Google AI (https://generativelanguage.googleapis.com/v1beta), and Local Ollama (http://localhost:11434/v1), DAVE pre-fills the default endpoint. Accept the pre-filled value unless you are using a non-standard deployment.
• For Azure OpenAI, OpenAI-Compatible, and Custom provider types, the Base URI is not pre-filled. Paste your endpoint URL into the Base URI field. Double-check the URL: a trailing slash or a missing path segment will cause all model fetch and agent call attempts to fail.

Enter the API key

• Paste your API key into the API Key field. The field is masked by default. Click the eye icon to reveal the value and confirm it pasted correctly, then click the eye icon again to re-mask it.
• Do not share the key with other team members via chat or email after this step. The key is stored encrypted on the server using AES-256-GCM authenticated encryption and is never returned through the API once saved. If a team member needs to verify the key, they must obtain a new one from the AI service's dashboard.
• For Local (Ollama) providers, the API Key field may be left blank if your Ollama instance does not require authentication. Check your Ollama configuration to confirm.

Configure the model catalog

• Choose a model catalog mode. Two options are available, toggled with the Fetch from API and JSON buttons:
  • Fetch from API (recommended for all types except Custom): click Fetch Models to retrieve the list of available models directly from the AI service. A successful fetch also confirms that your API key and Base URI are correct. Use the search box to filter the list, then check the models you want to include. Use All or None to bulk-select.
  • JSON (required for Custom providers; available for all types): paste or type a JSON array of model objects in the format [{ "id": "model-id", "name": "Display Name" }]. Use this mode if Fetch from API is unavailable for your provider type, or if you prefer to define the catalog manually.
• For Custom providers, Fetch from API is not available. You must use JSON mode and enter the model catalog manually.
• Curate the catalog: include only the models your team will actually use. A smaller catalog makes model selection in agent configuration faster and reduces the risk of agents being pointed at an unintended model.

Save and verify

• Click Create Provider. The new provider card appears in the provider list.
• If you used JSON mode for the model catalog, click the chevron on the new provider card to expand it and confirm the model list is correct.
• If you have not already done so via Fetch Models, verify connectivity now: click the pencil icon to open the edit form, switch to Fetch from API mode, and click Fetch Models. A successful fetch confirms the API key and Base URI are working. Click Cancel to close without saving changes.
• Confirm the provider card shows the correct Provider Type badge, Base URI, and model count.

Use this table to confirm the correct Base URI and model catalog mode for each provider type before filling in the form.

All seven provider types are supported in the same provider form. You can configure multiple providers of the same type, for example one OpenAI provider for production and a second for staging, by creating separate provider records with distinct names.

A configured provider does nothing on its own. It becomes useful the moment an agent references it. Here is what to do after your provider is saved and verified:

1. Create or update an agent. Navigate to Agents in the sidebar. In the agent configuration, select your new provider from the provider dropdown and choose a model from the catalog you just configured. The agent will use the provider's API key and Base URI at runtime.
2. Test the agent in a workflow instance. Create a test instance of a workflow that includes an agent interaction node using your new agent. Run the instance and confirm the agent executes successfully. A failed agent call at this stage usually means the API key is invalid or the model ID is incorrect: return to Admin, then API Management, then Providers and edit the provider.
3. Plan your key rotation schedule. Because all agents that reference this provider share a single API key, rotating the key here updates every agent at once. Decide how frequently you will rotate the key (monthly is a common baseline) and note the rotation date. When you rotate, paste the new key into the API Key field and click Update Provider. Use Fetch Models to confirm the new key works before saving.
4. Document the provider for your team. Add a note to your team's internal documentation recording the provider name, its purpose, the AI service account it connects to, and who is responsible for key rotation. The provider name is visible to all Admin users in the tenant, but the key itself is masked and encrypted: only the person who holds the service account credentials can rotate it.

Can I have more than one provider of the same type?
Yes. You can create multiple provider records of the same type, for example two OpenAI providers pointing at different API keys or rate-limit tiers. Give each a distinct, descriptive name so agents can be configured to use the correct one. This is a common pattern for separating production and staging environments.

What happens to my agents if I delete a provider?
Agents that reference the deleted provider will fail on their next invocation. Before deleting a provider, update every agent that uses it to point to a different provider. Then delete the original. The provider list page does not warn you about dependent agents before deletion, so check agent configurations first.

Who can see the API key after it is saved?
Nobody, including Admins, can retrieve the full API key through the DAVE interface or API once it is saved. The key is stored using AES-256-GCM authenticated encryption and is never returned through the API. The UI displays the field masked. If you need to verify a key, you must obtain it from the AI service's dashboard and re-enter it. This is by design: it means a compromised DAVE account does not expose your AI service credentials.

My Fetch Models attempt returned zero models. What do I do?
Three common causes: the API key does not have permission to list models; the Base URI is incorrect; or the AI service returned an empty list. For Azure OpenAI, confirm at least one model is deployed in your Azure resource. If Fetch Models consistently returns zero results, switch to JSON mode and enter the model IDs manually. You can find the correct model IDs in your AI service's documentation or dashboard.

Can non-Admin users see which providers are configured?
No. The Providers page at Admin, then API Management, then Providers is accessible only to users with the Admin role. Users with Create, Curate, Use, Reporting, or Financial roles cannot view the provider list, provider names, or any provider details. They can use agents that reference providers, but the underlying provider configuration is not visible to them.

Do I need a provider for every AI model I want to use?
No. One provider record can include multiple models in its catalog. You only need a separate provider record when you are connecting to a different API key, a different service endpoint, or a different AI platform. For example, if you want to use both GPT-4o and o3 from the same OpenAI account, add both models to a single OpenAI provider's catalog rather than creating two provider records.