Initial Configuration
After installing Simple Agent Designer and assigning permissions, follow these steps to configure the extension for first use.
Opening Setup
- Search for Agent Setup in the Tell Me search bar, or
- Navigate to the Agent Role Center and choose Setup from the Agent Designer section
The first time you open Setup, the extension automatically initializes default values and loads standard agents and workload types.
Step 1: Configure AI Providers
Before agents can process workloads, you need to configure at least one AI provider with a valid API key.
- On the Setup page, choose AI Providers from the action bar
- On first run, the extension automatically creates 11 provider entries — every supported provider except Azure Open AI (which you add manually because it requires your own deployment endpoint). You only need to enter an API key for the provider(s) you actually use.
- For each provider you want to use:
ChatGPT OpenAI
| Field | Value |
|---|---|
| Provider | ChatGPT OpenAI |
| Endpoint | https://api.openai.com/v1/chat/completions |
| Access Key | Your OpenAI API key |
To get your API key, click on the Access Key field — it will open the OpenAI API Keys page.
Anthropic Claude
| Field | Value |
|---|---|
| Provider | Anthropic |
| Endpoint | https://api.anthropic.com/v1/messages |
| Access Key | Your Anthropic API key |
To get your API key, click on the Access Key field — it will open the Anthropic Settings page.
Azure OpenAI
| Field | Value |
|---|---|
| Provider | Azure OpenAI |
| Endpoint | Your Azure OpenAI deployment endpoint URL |
| Access Key | Your Azure OpenAI API key |
Click on the Access Key field to open your Azure AI Foundry project.
LM Studio (Local Models)
| Field | Value |
|---|---|
| Provider | LM Studio |
| Endpoint | Your LM Studio server URL (e.g., http://localhost:1234/v1/chat/completions) |
| Access Key | Leave empty or enter the key configured in LM Studio |
Access keys are masked on the screen for security. The key is stored in the Business Central database.
The examples above are the most common providers. Simple Agent Designer supports 12 providers in total — including Google Gemini, xAI Grok, Mistral, DeepSeek, Moonshot Kimi, Alibaba Qwen, Zhipu GLM, and MiniMax. For per-provider endpoints and where to obtain each API key, see AI Provider Setup.
Step 2: Configure General Settings
On the Setup page, configure the general settings that apply to all workload processing:
| Field | Description | When to Set |
|---|---|---|
| General Journal Template | The journal template used when agents create journal entries | Required if agents create GL entries |
| General Journal Batch | The journal batch within the selected template | Required if agents create GL entries |
| No Series for G/L Documents | The number series for document numbering | Required if agents create GL entries |
These defaults can be overridden at the Workload Type level for specific workflows.
Step 3: Review System Prompts
System prompts define the standard messages used by the orchestration engine when communicating with AI providers.
- On the Setup page, choose System Prompts from the action bar
- Review the pre-loaded system prompts
The system prompts include:
| Prompt | Purpose |
|---|---|
| PleaseStart | Initial message sent to an agent when processing begins |
| AreYouComplete | Checks whether an agent has finished its task |
| NameAndJob | Introduces the agent's identity (name and role) |
| NotValidJson | Error message when AI returns invalid JSON |
| TaskComplete | Sent when a task is completed successfully |
| ReplyFrom | Formats replies between cooperating agents |
| Asking | AI setup instructions with company context |
| ALResult | Reports AL code execution results |
| ALCompletedWithoutErrors | Success message for AL execution |
| TaskFailed | Sent when a task fails |
System prompts support language-specific versions. The system first looks for a prompt matching the user's language, then falls back to the default (no language specified).
Editing System Prompts
- Select a system prompt in the list
- Choose Edit from the action bar
- The System Prompt Editor opens with a rich text editor for the instructions field
- Modify the prompt text as needed
- Close the editor to save
Most users can keep the default prompts. Customize them only if you need to change the orchestration behavior or add language-specific versions.
Step 4: Review Default Agents
Simple Agent Designer ships with pre-configured standard agents. Review and customize them:
- From the Role Center, choose AI Agents from the Agent Designer section
- Review the agent list showing Agent ID, Type, Title, AI Provider, LLM Model, and settings
- For each agent, verify:
- AI Provider — points to a provider you have configured with a valid key
- LLM Model — set to an appropriate model (e.g.,
gpt-5-mini,claude-opus-4-1) - Use Tools — enabled if the agent needs to interact with Business Central data
- General Agent — one agent should be marked as the default for the interactive chat
To edit an agent's instructions, select it and choose Edit Instructions from the action bar.
Step 5: Review Workload Types
Standard workload types are loaded during initialization:
- From the Role Center, choose Types from the Work section
- Review the available workload types
- Each type defines:
- Type — unique identifier
- Description — what this type of workload does
- Initial Worker — which agent processes this type
- Operation — PayloadBased or DataBased
- Excel Splitter — whether to split Excel file uploads
You can use the standard types as-is, modify them, or create your own.
Step 6: Test Your Setup
Verify everything works by creating a simple test workload:
- From the Role Center, go to Workloads
- Choose New Workload
- Select a workload type
- Enter a simple task description (e.g., "What is the current date and time?")
- Choose Process
- Monitor the execution in the Process Chat page
- Verify the agent responds correctly
If the test succeeds, your setup is complete and you are ready to start building real workloads.
Quick Setup Checklist
- Extension installed and AI Accountant permission set assigned
- License registered (required for production; free in a SaaS sandbox — see Registration)
- 11 providers pre-seeded on first run — API key added for the provider(s) you use
- General Journal settings configured (if using journal entry agents)
- Default agents reviewed and pointed to active AI providers (default model e.g.
gpt-5-mini) - Workload types reviewed
- Test workload processed successfully
Note: Agents can now do more than analyze text. They can run read-only SQL queries against Business Central tables, send email notifications to a workload's configured recipients, and generate Excel (.xlsx) files — all through built-in tools. See Key Features for details.