Agent playground overview (preview)

[This article is prerelease documentation and is subject to change.]

The Agent playground is a prototyping environment that allows developers and partners to experiment with AI capabilities in Business Central by building playground agents. It's available in sandbox environments and designed for testing and learning purposes, providing a safe environment to explore AI functionalities without affecting production data.

Purpose and audience

The Agent playground is primarily designed for teams who want to prototype AI-powered capabilities by configuring playground agents:

• Product owners who want to prototype AI-powered features before building them in AL extensions
• Domain experts evaluating AI integration possibilities for their Business Central solutions
• Consultants exploring automation scenarios with AI assistance for client solutions
• Developers validating AI use cases and gathering requirements for custom implementations

The playground helps you understand how AI can interact with Business Central user interface and data, allowing you to test various AI functionalities such as automating tasks, answering questions, and integrating AI into workflows through playground agents.

When to use and when not to use

When to use the Agent playground

Use the Agent playground when you want to:

• Prototype AI features before investing in AL development
• Test AI interactions with Business Central UI and data
• Validate use cases and gather requirements for AI-powered solutions
• Experiment with prompts and AI behavior in a controlled environment
• Demonstrate AI capabilities to stakeholders or customers
• Learn AI integration patterns for Business Central

When not to use the Agent playground

Don't use the Agent playground when:

• Production environments are involved (sandbox only)
• Compliance or security evaluation is the primary goal
• Final implementation is ready (graduate to AL extension instead)
• End-user training is the main objective (use production-ready Copilot features)

Agent lifecycle summary

A typical playground agent follows this lifecycle:

1. Create - Set up the agent with basic identity and profile
2. Configure - Define instructions and permissions
3. Activate - Enable the agent for testing, add tasks
4. Iterate - Test, refine instructions, and adjust permissions
5. Graduate - Export learnings to build production solutions using AL extensions
6. Clean up - Deactivate and remove the prototype agent

Each phase has specific requirements and limitations that guide your development process.

High-level concepts

Understanding the following key concepts helps you work effectively with playground agents.

Prerequisites

Before you can use the Agent playground, ensure that you have the following prerequisites in place:

• Enable the Custom Agent capability in Business Central. Learn more in Configure Copilot and agent capabilities.
• Appropriate permissions to access and use the Agent playground in Business Central.
  • "AGENT - ADMIN" gives permission to work with the playground.
  • "AGENT - DIAGNOSTICS" gives access to extra troubleshooting data.
• A sandbox environment for testing and experimentation from version 27.2 (can include customer data such as sandbox copies of production data).
• Billing setup for agent capabilities in your tenant. Learn more in Manage consumption billing for agents (preview).

Instructions and tasks

Instructions are natural language descriptions that define your agent's purpose and behavior. Think of them as detailed prompts that guide how the agent interprets requests and interacts with the Business Central interface. Well-written instructions include context, clear expectations, and specific guidance about the agent's role. Learn more about instructions in Write effective instructions for a playground agent (preview).

Tasks are a way to trigger the agent. Tasks can only be added when the agent is active, and they help structure how the agent responds to different types of requests. Learn more about tasks in Run a playground agent (preview).

Supported languages

This feature currently supports only English as language. While this feature can be used in other languages, it might not function as intended. Language quality might vary based on the user's interaction or system settings, which might affect accuracy and the user experience.

Learn more about geographic and language availability at Copilot international availability.

Permissions and profiles

The agent operates with explicit permission sets that control what data and actions it can access. This follows the principle of least privilege—grant only the permissions necessary for the agent's intended functionality. Permission sets can only be modified when the agent is deactivated in the Agent Card. From the agent's setup page, permissions can be modified while the agent is active, but the changes only affect new tasks or tasks, which are paused and then restarted.

The agent is assigned to a profile (role) that determines which UI elements it can see and interact with. To reduce contextual noise and increase accuracy, it's highly recommended to create custom profiles specifically for agents, using page customization properties to control visibility of actions, layouts, views, and operations like insert/delete.

Learn more in Set up playground agent permissions and profiles (preview).

Limitations (preview)

As a preview feature, the Agent playground has several important limitations:

Environment restrictions

• Sandbox only: Available exclusively in sandbox environments—never in production
• Sandbox data: Can be used with customer data in sandbox environments (including sandbox copies of production data)
• Preview functionality: Features and behavior might change without notice

Operational constraints

• State dependencies: Permission changes require deactivation when on the Agent Card; task changes require activation
• Permission inheritance: Agents can't have more permissions than their creating user
• UI interaction limits: Agents see only what their assigned profile allows
• Single list per page: Agents can only interact with one list on any given page. If a page has multiple sublists (for example, sales lines and notes in a factbox), the agent can only access one of them
• No Tell me: Agents can't use Tell me to find specific information. Navigation is limited to available actions and links

Development considerations

• Prototype only: Not suitable for production implementations or compliance evaluation
• Manual telemetry: No automatic logging—capture prompts and responses manually during testing
• Temporary nature: Designed for experimentation, not long-term solutions

Getting started

Ready to try the Agent playground? The process involves enabling the feature, creating your playground agent, and configuring it for your specific use case. Learn more in Create and activate a playground agent (preview).

Related information

Overview (preview)
Create and activate (preview)
Write effective instructions (preview)
Instruction keywords (preview)
Integrate with the Tasks AL API (preview)
Set up permissions and profiles (preview)
Run a playground agent (preview)
Iterate and manage (preview)
Best practices (preview)
Create a Sales Validation Agent (preview)
Transparency note: Business Central agent playground