Simple Agent Designer Documentation
Index

Key Features

Simple Agent Designer provides a comprehensive set of features for building, running, and managing AI agents within Microsoft Dynamics 365 Business Central. This page summarizes the main capabilities of the extension.

Multi-Provider AI Support

Simple Agent Designer supports 12 AI providers, giving you flexibility in choosing the right model for each agent's task.

Provider Description Use Case
Azure Open AI Microsoft-hosted OpenAI models Enterprise environments requiring Azure compliance
ChatGPT Open AI OpenAI's cloud-hosted models (GPT-5, GPT-4.1, etc.) General-purpose AI tasks, strong reasoning
Anthropic (Claude) Anthropic's Claude models (Claude Opus, Sonnet, Haiku) Complex analysis, nuanced understanding
LM Studio Locally hosted open-source models Privacy-sensitive workloads, offline operation
Google Gemini Google's Gemini models Multimodal tasks, large context
xAI Grok xAI's Grok models General-purpose reasoning
Mistral Mistral AI models Cost-effective general tasks
DeepSeek DeepSeek models Reasoning and coding tasks
Moonshot Kimi Moonshot AI's Kimi models Long-context processing
Alibaba Qwen Alibaba's Qwen models Multilingual tasks
Zhipu GLM Zhipu AI's GLM models General-purpose tasks
MiniMax MiniMax models General-purpose tasks

Each agent can be independently configured with its own provider and model, allowing you to optimize cost and performance across your workflow. For per-provider endpoint and API key details, see AI Provider Setup.

Flexible Agent Design

Agents (Workers) are the heart of Simple Agent Designer. You can create specialized agents for different business functions:

  • General AI agents — for open-ended analysis, summarization, and decision-making
  • Data query agents — for looking up and retrieving Business Central data
  • Journal entry agents — for creating general ledger journal entries
  • Sales document agents — for creating sales quotes, orders, invoices, and credit memos
  • Purchase document agents — for creating purchase quotes, orders, invoices, and credit memos
  • AL runner agents — for executing AL code snippets generated by other agents
  • Data modifier agents — for updating existing Business Central records

Each agent receives custom instructions that define its role, behavior, and output format.

Workload Processing System

The workload system provides a structured approach to task management:

  • Workload Cards — individual task definitions with rich text descriptions and file attachments
  • Workload Types — reusable templates that define processing rules, data sources, and agent assignments
  • Batch Processing — process all pending workloads with a single action
  • Status Tracking — monitor workload progress (Not Started, Started, Success, Fail)
  • Document Linking — automatically track documents created by workload processing

Data Splitting for Bulk Operations

When you need to process large datasets, Simple Agent Designer can automatically split them into individual workloads:

Excel Splitting

  • Upload an Excel file and have each row processed as a separate workload
  • Supports files with or without header rows
  • Automatic title generation using AI summarization
  • Duplicate detection to prevent reprocessing

Table Splitting

  • Select any Business Central table as a data source
  • Apply filters to target specific records
  • Incremental processing with "since last run" filtering
  • Each matching record becomes its own workload with full record data as context

Tool and Function Calling

Agents can interact with Business Central through a powerful tool system:

Built-in Tools

  • Data operations — look up, create, and update records in any Business Central table
  • Document creation — create journal lines, sales documents, and purchase documents
  • Report execution — list, configure, and run Business Central reports
  • Dimension management — resolve dimension set combinations

Custom Tools

  • Design your own tools using the built-in Tool Designer
  • Define tool names, descriptions, and parameter schemas
  • Tools are dynamically registered and available to all agents with tools enabled

Agent Cooperation

Multiple agents can collaborate on complex tasks through the orchestration engine:

  • Agents can reference other agents in their responses to delegate subtasks
  • The orchestration engine routes messages between agents automatically
  • Results from specialized agents are returned to the requesting agent
  • Worker chains can involve any number of agents working together

Interactive Agent Chat

Simple Agent Designer provides two ways to chat with your AI agents in real time:

Role Center Chat

The Agent Based Accounting Role Center includes a built-in chat part for quick agent interactions alongside your work tracking.

Ad Hoc Agent

A dedicated full-screen chat page with its own navigation, giving you direct access to agent designer, tools, skills, and workload management from a single workspace.

Chat Capabilities

Both chat interfaces support:

  • Real-time conversation with your default agent
  • Automatic tool execution during chat
  • Interrupt — send a new message while the agent is executing tools to redirect the conversation
  • Saved conversations — chat history is auto-saved and can be resumed later via the Resume Conversation and New Conversation actions

In addition, the Ad Hoc Agent page supports drag-and-drop file upload — drop images, PDFs, Excel, or CSV files directly into the chat. This is not available in the Role Center Interactive Agent part; use the Ad Hoc Agent when you need to upload files in chat. The Ad Hoc Agent also provides a Clear Agent action to delete agent work and start fresh.

Skills System

Skills are reusable knowledge entries that teach your agents how to perform specific tasks:

  • Create skills manually — write step-by-step procedures for your agents to follow
  • Import from web articles — enter a URL (e.g., a Microsoft Learn page) and the system automatically extracts knowledge into a skill
  • Agent access — agents retrieve skills using the get_skill tool during processing
  • Build up a knowledge base that makes your agents more capable over time

Role Center Integration

A dedicated role center ("Agent Based Accounting") provides a centralized workspace:

  • Agent Work tracking — see recent work performed by agents
  • Quick access — links to workloads, types, agents, tools, skills, and setup
  • New Work wizard — guided workload creation with file upload or text description, choice of workload type, and foreground or background processing
  • Automation — inbound workload support for external integrations

Standard Page Extensions

Simple Agent Designer extends standard Business Central pages to surface AI-generated content:

  • General Journal — AI Accountant factbox showing related workload information
  • Payment Journal — AI Accountant factbox for payment context
  • Cash Receipt Journal — AI Accountant factbox for receipt context
  • Customer Card — Bank statement name field for enhanced transaction matching

Import and Export

Share agent configurations across environments:

  • Export Agents and Types — create a JSON package of all agents and workload types
  • Import Agents and Types — restore from a previously exported package
  • Import Standard Agents and Types — reload default configurations from the built-in resource

Inbound Workloads and Automation

External applications can submit work to Simple Agent Designer through an API:

  • API endpoint for receiving workloads from external systems
  • Inbound Request List to review and process incoming requests
  • Choose between manual review or automatic processing
  • Integrates with the Mobile App and any external system

Mobile App Integration

A companion mobile app connects to Business Central for on-the-go work submission:

  • QR code setup — scan a QR code to configure the mobile app connection
  • Mark workload types as available in the mobile app
  • Enable photo capture for image-based workloads
  • Enable file upload for document processing
  • Provide text instructions for guided data entry