Infinity Run MCP architecture and governance

Pega Infinity™ Run MCP is the run-time implementation of the MCP in Pega Infinity™. It supports governed interactions between Pega applications and MCP-compatible systems. External AI agents can discover and invoke Pega business capabilities as MCP tools. As part of an agent or Case workflow, Pega applications can connect to external MCP servers and invoke their available tools. 

Infinity Run MCP supports these run-time interactions: 

• External AI agents can discover available Pega business capabilities. 
• External AI agents can invoke approved Pega workflows and business operations. 
• Pega runs and governs the associated workflows and Case processes. 
• Existing security, authorization, governance, compliance, and audit controls continue to apply. 
• Pega returns the interaction results to the initiating system. 
• Note: The AI agent interprets the intent, while Pega governs and runs the business process. 

Examples include an external AI agent identifying a customer’s intent to file an insurance claim, dispute a transaction, request leave, or update account information. Instead of embedding the required business logic in the AI agent, the agent invokes the appropriate Pega capability through MCP. Pega then runs the workflow with the same validations, routing, approvals, security controls, and compliance requirements that apply when the Pega Portal initiates the process. 

At runtime, an external AI assistant connects to an MCP endpoint exposed by a Pega application. The external client discovers the tools exposed by the Service MCP Rule and can select and invoke a tool based on the request and the client’s own reasoning and configuration. Pega runs the underlying workflow and returns the result to the agent. 

A common point of confusion is the relationship between Infinity Run MCP and Service MCP. 

Infinity Run Model Context Protocol (MCP) supports run-time interactions between Pega Platform and external MCP-enabled systems. Authorized external agents can invoke Pega tools and Case Types. 

Service MCP is the Rule implementation mechanism that makes Infinity Run MCP possible. A Service MCP Rule creates an MCP endpoint, defines which tools and Case Types are exposed, and controls how external agents connect to and invoke those capabilities. 

While Infinity Run MCP focuses on exposing governed Pega capabilities to external AI agents, the broader MCP runtime architecture consists of four related components. 

Service MCP Rule: The Service MCP Rule debuted in Pega Infinity  26 and serves as the primary building block for Infinity Run MCP. The rule exposes selected Case Types, workflows, and business functions as MCP tools through a secure MCP endpoint. External AI agents use this endpoint to discover and run these tools. 

MCP Tools: MCP tools represent the business capabilities that Service MCP exposes. These tools can start workflows, process assignments, retrieve data, invoke automations, and perform other approved business operations. Pega also provides pxCommon, an out-of-the-box MCP service that exposes a standard set of MCP tools with minimal configuration. Its MCP endpoint is created automatically and exposes a predefined set of tools that external agents use to discover Case Types and interact with workflows with minimal configuration. Organizations can start with these capabilities and then create custom Service MCP Rules for business-specific requirements. 

Connect MCP Rule: The Connect MCP rule provides the opposite interaction pattern. Instead of exposing Pega capabilities, a Pega Agent uses the Connect MCP rule to connect to an external MCP server and invoke external MCP tools. In this model, Pega acts as an MCP client. 

MCP Connect Step: The MCP Connect step integrates external MCP tools with case workflows. A case process can invoke an external MCP tool at a specific point in the workflow without a separate Pega Agent. 

A key principle of Infinity Run MCP is that governance does not move into the AI agent. Instead: 

• Security remains inside Pega. 
• Authorization remains inside Pega. 
• Workflow routing remains inside Pega. 
• Compliance controls remain inside Pega. 
• Business rules remain inside Pega. 
• Auditability remains inside Pega. 

The external agent simply invokes a tool. Pega remains responsible for execution and governance. MCP access remains subject to Pega authentication, authorization, and auditability. Architects must configure the OAuth client, operator mapping, privileges, and exposed tools so that each client receives only the required access. 

Before Infinity Run MCP, organizations typically had to build separate integrations for each chatbot, copilot, voice assistant, and AI platform. This approach often resulted in duplicate business logic, inconsistent  behavior, and additional maintenance effort. 

With Infinity Run MCP, organizations can publish workflows once and reuse them across multiple AI agents and channels. A single governed process can support customer-facing assistants, employee copilots, and external agent platforms without separate workflow implementations. 

Infinity Run MCP supports integration and provides a standardized, governed approach to runtime interactions between Pega Platform applications and MCP-compatible systems. Organizations can use Infinity Run MCP to expose enterprise workflows to external AI agents and integrate external capabilities into Pega agents and Case workflows while maintaining enterprise control. 

As organizations adopt customer-facing and employee-facing AI agents, Lead System Architects (LSAs) need to maintain consistent business processes across all interaction channels. Infinity Run MCP helps LSAs centralize workflow orchestration, approvals, checks that support compliance strategies, security controls, and audit trails in Pega Platform while supporting governed interactions with external systems. 

Centralizing business processes and governance in Pega Platform, instead of distributing these functions across multiple AI assistants, reduces integration complexity, improves process consistency, and maintains enterprise control. AI agents provide the interaction and reasoning layers, while Pega Platform coordinates and governs the workflows and external capabilities that produce business outcomes. 

Check your knowledge with the following interaction: