Simple Agent Designer

Setup AI Providers

Connect AI Providers to Simple Agent Designer

This guide walks you through adding AI providers to the Simple Agent Designer and assigning a provider/model to each agent.

At a glance

  • Supported providers: OpenAI (ChatGPT), Anthropic (Claude), Azure OpenAI

  • Business Central Copilot subscriptions aren’t supported for this app (they don’t expose features we need)

  • You can mix and match providers and models per agent

  • Azure OpenAI uses your own endpoint + deployment; the in-app Model field is ignored for Azure

1) Open the AI Providers page

  1. In Simple Agent Designer, go to Setup → AI Providers.

  2. You’ll see any providers you’ve already configured.

2) Add a provider

You can add any (or all) of the three supported providers. Each needs an API key. The app includes an AssistEdit (little icon next to the Access Key field) that opens the correct provider page for generating your key.

A. OpenAI (ChatGPT) — easiest setup

  1. In AI Providers, add a new line and choose OpenAI (ChatGPT).

  2. Click AssistEdit next to Access Key.

    • If you don’t have an OpenAI account, sign up when prompted.

    • Create a new API key on the OpenAI platform.

  3. Paste the key into Access Key and save.

You’ll assign a model and agent usage in step 4.

B. Anthropic (Claude)

  1. In AI Providers, add a new line and choose Anthropic (Claude).

  2. Click AssistEdit next to Access Key.

    • Sign in (or sign up) to the Anthropic console.

    • Create a new API key.

  3. Paste the key into Access Key and save.

C. Azure OpenAI — needs your endpoint + deployment

Azure OpenAI requires a deployed model (e.g., GPT-4.1) in your Azure subscription.

  1. In AI Providers, add a new line and choose Azure OpenAI.

  2. Click AssistEdit to open Azure AI Foundry (Azure OpenAI Studio).

  3. Create (or locate) a Deployment for the chat model you want to use.

  4. In the deployment page, click View code under Chat to find:

    • Your Endpoint URL (includes your resource name, e.g., https://<your-resource>.openai.azure.com/...)

    • Your Deployment name (e.g., deployment-1)

    • Your API Key

  5. Copy these values into the provider line in Simple Agent Designer:

    • Endpoint URL → paste your Azure endpoint

    • Deployment → paste your deployment name

    • Access Key → paste your Azure API key

  6. Save.

Important: For Azure OpenAI, the Model field in the Agents page is ignored—the model comes from your Azure deployment.

3) (Optional) Check usage & costs on provider consoles

Each provider has a usage dashboard:

  • OpenAI: total tokens, requests, and spend

  • Anthropic: similar usage metrics

  • Azure OpenAI: resource-level quotas and usage

Use these to keep an eye on consumption (e.g., choosing “mini” models for cost-sensitive agents).

4) Assign a provider and model per agent

Configuring providers isn’t enough—each Agent must be told which provider/model to use.

  1. Go to Agents.

  2. Open an agent and set:

    • Provider: choose OpenAI, Anthropic, or Azure OpenAI

    • Model:

      • OpenAI/Anthropic: pick a model (e.g., GPT-4.1, or Mini for lower cost)

      • Azure OpenAI: Model is ignored (your Azure Deployment defines the model)

  3. Save.

You can mix and match. Example: use Claude for long reasoning tasks and GPT-4.1 mini for quick utilities. Different agents can target different providers/models.

Tips & notes

  • Why not Business Central Copilot? The features we require aren’t available via the Copilot subscription exposed through BC, so bring your own provider subscription.

  • Key security: API keys are secret—store them only in the Access Key field. You won’t be able to view existing keys in full later (by design).

  • Costs: “Mini” models are cheaper and often sufficient for lightweight tasks. Start conservative; scale up as needed.

  • Azure specifics: If an agent is set to Azure OpenAI but “doesn’t respond,” double-check:

    • The Endpoint URL is correct for your resource

    • The Deployment name matches your Azure deployment exactly

    • Your Azure key is valid and you have quota

Troubleshooting

  • 401/Forbidden / Invalid key

    • Regenerate the key in the provider console, update the Access Key, and save.

  • 404 / Model not found (Azure)

    • The Deployment name doesn’t match, or you haven’t deployed that model. Verify in Azure AI Foundry → Deployments.

  • Timeouts / Slow responses

    • Try a smaller model (e.g., Mini) or a different provider; check provider status pages.

  • Agent ignores Model selection (Azure)

    • Expected. Azure uses the model in your Deployment, not the Agents page.

FAQ

Can I use multiple providers at once?
Yes—configure several providers and select different ones per agent.

Which model should I choose?
Start with GPT-4.1 (general strong default) or GPT-4.1-mini (cost-efficient). For very long tool-use/analysis tasks, try Claude variants.

Do I need to do anything after adding providers?
Yes—assign a provider (and model) to each agent in the Agents page.

Quick checklist

  • Added providers in Setup → AI Providers

  • Generated and pasted API Keys (and Endpoint/Deployment for Azure)

  • For each agent: chose Provider (and Model, except on Azure)

  • Verified usage/costs in provider consoles

You’re all set! Try it in the included sandbox and see what works best for your scenarios