Center-out Agent Architecture

Interaction channels, including web applications, mobile apps, chatbots, and copilots, increasingly use AI agents. Back-end systems can also use agents to interact with customers, employees, enterprise data, workflows, and other agents. 

If each agent maintains separate business rules and decision logic, agents can produce inconsistent outcomes. Distributing this logic can also fragment governance, make compliance more difficult to manage, and reduce visibility into how agents produce business outcomes. 

Successful adoption therefore depends on more than the reasoning capabilities of individual agents. Organizations also need to orchestrate and govern agent activity and align it with business objectives, policies, and approved processes.

Pega's Center-out Agent Architecture organizes AI, workflows, and decisioning around customer and business outcomes. 

Reusable business capabilities form the governed center of the architecture. These capabilities include workflows, business Rules, and decisioning implemented with Pega Platform™. Interaction channels, AI agents, enterprise systems, data sources, and processing can remain distributed. 

This separation enables channels and agents to invoke shared business capabilities without embedding copies of the underlying business logic. 

The following figure illustrates this center-out relationship between distributed components and governed business capabilities in Pega Platform. 

Three steps to a Center-out architecture

The following conceptual steps illustrate how the Center-out approach transforms fragmented interactions into coordinated journeys. 

1. Define the journey: Design journeys around customer and business outcomes. Use centrally managed workflows, business Rules, and decisioning to define how the organization handles each Outcome. Pega Blueprint™ uses generative AI to suggest foundational application assets, including Workflows, Workflow Details, Workflow Data, and Personas. Teams review and refine these assets to align the design with business requirements and intended Outcomes. 
2. Connect interaction channels: Use  Pega Application Agent™ endpoints to integrate conversational Pega Agents with supported Channel experiences. Use Pega Digital Experience API (Pega DX API) to build Channel-specific experiences that interact with Pega Cases, Assignments, data, and Agents. This approach enables Channels to reuse governed workflows while retaining user experiences suited to each Channel.
3. Access enterprise data: Use Pega Live Data™ to abstract access to data in legacy and modern systems. Applications reference business data without coupling workflows to the technical details of the underlying data sources. Live Data can retrieve information on demand, reuse cached information, and combine information from multiple sources according to the application’s data configuration. 

Pega Platform can make governed business capabilities available to AI agents through open interoperability protocols. This approach enables agents to use existing workflows, tools, and business capabilities while Pega continues to manage the underlying processes and business logic. 

Model Context Protocol (MCP) 

Pega Platform supports MCP to connect Pega capabilities with external AI systems. Depending on the integration direction and use case, Pega provides several MCP mechanisms: 

• Service MCP Rule: Exposes selected Pega Platform capabilities as tools that authorized external MCP clients can invoke. 
• Connect MCP Rule: Connects a Pega Agent to an external MCP server so that the Agent can invoke tools provided by that server. 
• pxCommon MCP tools: Provide predefined tools that can give external agents access to supported Pega application capabilities. 
• MCP Connect Step: Enables a workflow to invoke a specific tool from an external MCP server through an Connect MCP Rule, without requiring a Pega Agent. 

Agent2Agent (A2A) protocol 

Pega Platform supports A2A to enable communication between Pega Agents and external agents for completing goal-oriented tasks. Pega supports A2A interactions in several ways: 

• Pega Agent to external agent: A Pega Agent can connect to an external agent to use its capabilities as part of completing a task. 
• External agent to Pega Agent: An external agent can connect to an Agent created in Pega Platform. 
• Agent discovery and collaboration: In Pega Agentic Process Fabric™, Intelligent Registry facilitates discovery of agents and workflows and supports A2A-based agent collaboration. 
• Agent selection and task delegation: Fabric Control Agent identifies and selects an appropriate agent from Intelligent Registry to facilitate task delegation and completion. 

At the enterprise level, Pega Agentic Process Fabric provides orchestration across connected agents and workflows. Its Intelligent Registry maintains information about registered resources so that agents can discover appropriate capabilities and delegate work  while maintaining governance and operational visibility. 

Business outcomes 

Organizations can use the Center-out approach to: 

• Centrally manage reusable business logic while keeping channels, data sources, systems, and processing distributed. 
• Produce more consistent outcomes across channels, systems, and AI agents. 
• Apply enterprise policies and governance controls to agent-enabled business processes. 
• Support transparency and auditability and provide mechanisms for implementing regulatory compliance controls. 
• Apply changes more consistently by updating reusable business logic instead of duplicating changes across channels and systems. 
• Coordinate customer and employee journeys across the enterprise. 

With the Center-out approach, AI agents can reason about and perform delegated tasks, while Pega Platform applies centrally managed workflows, business Rules, decisioning, and governance controls to the associated business processes. 
 

Check your knowledge with the following interaction: