What is the Model Context Protocol (MCP) for Contact Centers in 2026?

The Model Context Protocol (MCP) is an open-source standard, originally created by Anthropic in November 2024, that provides a universal way for AI models to connect with external tools, data sources, and applications. Often described as the “USB-C of enterprise AI,” MCP standardizes how AI systems interact with enterprise software through a structured JSON-RPC 2.0 communication layer, delivering flexible integration and API extensibility that eliminates the need for custom-coded, point-to-point connectors.

Before MCP, connecting AI models to enterprise tools such as CRMs, ticketing systems, ERP platforms, and knowledge bases required a unique integration for every combination. An organization with 10 AI applications and 100 business tools could face up to 1,000 distinct integration pathways. MCP converts this exponential problem into a simple additive formula: each tool needs one MCP server, and each AI model needs one MCP client. The result is immediate, plug-and-play interoperability across the entire software ecosystem, enabling a true multi-vendor AI strategy where enterprises choose the best model for each task without rebuilding their infrastructure.

In December 2025, Anthropic formally donated MCP to the newly established Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation. Co-founded by Anthropic, OpenAI, and Block, with Platinum membership from AWS, Google, Microsoft, Bloomberg, and Cloudflare, the AAIF ensures vendor-neutral governance and long-term ecosystem stability. This open standards foundation guarantees that MCP’s evolution is driven by community consensus, not proprietary roadmaps, giving enterprises confidence in a future-proof AI architecture.

For customer experience leaders, MCP is the architectural enabler that transforms the contact center from a reactive routing engine into a proactive, AI-powered orchestration platform that delivers a unified experience across every channel and system. Traditional contact centers suffer from a persistent problem: agents toggle between disconnected applications to piece together a customer’s history, preferences, and current situation. A customer calling about a billing issue might have relevant information spread across a CRM, a ticketing system, an EHR (in healthcare), and a knowledge base. Surfacing all of that context in real time previously required expensive, brittle, point-to-point integrations that most organizations could never fully build or maintain. MCP makes deep AI integrations across these systems not only possible but practical.

MCP eliminates this context blind spot. When a customer interaction begins, the platform’s embedded AI dynamically discovers the necessary MCP tools and orchestrates data retrieval in real time. The AI does not need to understand the schema of every external system natively. It simply connects through the standardized protocol, pulling exactly the context it needs to personalize the interaction, delivering operational efficiency gains that compound across every customer touchpoint.

The first generation of AI in the contact center focused almost entirely on deflection: routing customers away from human agents to reduce costs. While this approach delivered short-term savings, it often degraded brand loyalty and customer satisfaction, particularly for complex or emotionally sensitive interactions. MCP enables a fundamentally different model built on AI-driven workflow orchestration. Instead of replacing human agents, AI becomes a continuous orchestration engine that provides real-time, context-rich support during live interactions. The AI quietly executes database lookups, pulls contextual history, surfaces actionable intelligence, and drafts responses for human review. Human agents remain in control but are now empowered agents with streamlined workflows and situational awareness that was previously impossible, delivering AI-driven efficiency without sacrificing the human connection.

Because MCP standardizes tool interaction and data flow, it dramatically lowers the barrier to building sophisticated customer workflows and enables continuous AI evolution without re-architecting existing systems. Contact center administrators and business users can configure dynamic, context-aware journeys without requiring deep software engineering expertise. Business users can select which AI models to use, define their own rules, and compose agentic workflows that adapt in real time to customer needs. This democratization of orchestration delivers a modern AI experience that accelerates time-to-value and enables organizations to iterate on CX strategies at the speed of business rather than the speed of IT, turning native AI innovation into an ongoing operational advantage rather than a one-time deployment event.

MCP is built on a three-layer client, model, and server architecture:

The Client gathers context, including user identity, session history, role, current task, and environmental variables, and sends it to the AI model. In a contact center, this is the agent desktop, IVR system, or customer-facing chat interface.

The Model (such as Google Gemini, Anthropic Claude, OpenAI GPT, or open-weight models like LLaMA or Mistral) receives the context and a list of available tools or actions. It determines which tools to invoke and what data to request.

The Server hosts the tools, APIs, databases, and data sources and executes any actions the model requests. In a contact center environment, MCP servers connect to CRMs (Salesforce, Microsoft Dynamics), knowledge bases, EHR systems, policy databases, scheduling platforms, and workforce management tools.

All communication follows structured schemas defined in JSON or YAML, making every interaction interpretable, traceable, and secure. This standardized intermediary layer converts the complex N x M integration problem (where N tools multiplied by M models equals exponential complexity) into a simple N + M solution, enabling complex environment orchestration and seamless integration at enterprise scale.

Understanding which platforms have adopted MCP and how they implement it is essential for CX decision-makers evaluating their next contact center architecture.

The competitive takeaway is clear: the race in CX is no longer about who has the best proprietary chatbot. For AI for large enterprises, it is about which vendor provides the most flexible, secure, and open orchestration engine for autonomous agentic interactions with the governance and scalability that complex organizations require.

The question facing every enterprise is no longer “Do we adopt AI?” That decision has already been made. The question now is: Who controls the intelligence of your business?

As Avaya’s approach demonstrates, AI openness built on protocols like MCP restores control to the enterprise. The five pillars of open AI architecture are: open protocols (standardized communication rules like MCP), model neutrality (the ability to swap LLMs without re-architecting, enabling multi-AI provider support), agent-agnostic orchestration (support for agents from any vendor), interoperable APIs (universal standards like JSON-RPC 2.0 with full API extensibility), and context continuity (shared memory across agents and systems delivering integrated AI capabilities that compound over time).

Gartner’s 2025 Innovation Insight report identified MCP as fundamental to the future of AI connectivity and enterprise architecture. According to Gartner’s projections, 75% of API gateway vendors and 50% of Integration Platform as a Service (iPaaS) vendors will natively support MCP features in 2026. Gartner anticipates that 33% of all enterprise software will feature agentic retrieval-augmented generation (RAG) capabilities by 2028, driven almost entirely by standardized interoperability protocols like MCP. That figure was less than 1% in 2025.

Boston Consulting Group has characterized MCP as a “deceptively simple idea with outsized implications,” noting that without MCP, integration complexity rises quadratically as AI agents spread throughout organizations. With MCP, integration effort increases only linearly, a critical efficiency gain for enterprise-scale deployments.

Research from MIT Sloan found that proprietary models cost six times more per million tokens than open-weight models while delivering only marginal accuracy improvements over time, with estimated global savings of $24.8 billion annually from shifting workloads where appropriate. For enterprises evaluating total cost of ownership (TCO), MCP delivers cost-effective AI by decoupling model usage from vendor fees and enabling predictable costs through standardized, reusable integrations. The value beyond price is equally significant: organizations gain the agility to adopt breakthrough models as they emerge without re-engineering their entire CX stack.

“Nearly 150 organizations joining the AAIF in its early days is a strong signal that agentic AI is shifting from experimentation to real-world deployment. The infrastructure for autonomous systems must be open, interoperable, and community-governed.”

— Jim Zemlin, Executive Director, Linux Foundation
 

Healthcare: How Does MCP Enable a Hybrid Healthcare Agent?

A healthcare provider can use MCP to coordinate AI-assisted medical history summarization with human nurse escalation for high-priority cases, combining enterprise-grade resilience with mission-critical continuity for patient care workflows. Each agent can access the patient’s case file in real time via a central memory layer using open standards and FHIR-compliant MCP servers. No data is duplicated or lost in handoffs. Specialized extensions such as the Healthcare Model Context Protocol (HMCP) enable AI to securely interact with clinical data systems while maintaining compliance with HIPAA, GDPR, and FHIR regulations, reflecting the robust data protection that regulated industries demand. The AI can autonomously draft prescriptions and schedule follow-ups based on real-time electronic health records, subject to clinician review. Expected impact: faster resolution, fewer callbacks, higher staff satisfaction.

Financial Services: How Does MCP Power a Compliance-Governed Copilot?

A global bank can use MCP to orchestrate a private, sovereign-hosted LLM trained on regulatory policy alongside a policy database agent providing explainable AI decision trees, a Specialized Action Model for transaction anomaly detection, and a human compliance officer equipped with a secure AI copilot. Multinational banks are deploying MCP to connect AI agents with real-time market data and transactional ledgers. By eliminating the latency of traditional batch processing, agentic AI can instantly assess transaction risk, detect fraud patterns, and trigger step-up authentication during live customer interactions, delivering AI-driven compliance that satisfies regulators while accelerating service. New models can be layered in over time without re-architecting the system. Expected impact: faster decisions, zero governance gaps in regulatory audits.

Retail: How Does MCP Unify an Omnichannel Customer Journey?

A major retailer can harmonize voice IVR agents on AWS Lex, chatbots on Meta’s LLaMA 3, store associate copilots, and loyalty agents using historical data. Unlike platforms that silo data by channel, an open MCP architecture ensures context persists across all systems. A customer who starts a return online can walk into the store and be greeted by an associate already briefed on the interaction. The MCP layer also enables real-time product recommendation engines, inventory lookup, and personalized promotional offers, all orchestrated through a single protocol rather than dozens of brittle custom integrations.

Government: How Does MCP Support Sovereign AI Deployment?

Government agencies with strict data sovereignty requirements can deploy MCP across on-premises and sovereign cloud environments, leveraging on-premises integration alongside strategic cloud adoption for a controlled transition that delivers enterprise modernization with minimal disruption. Multilingual orchestration supports constituent service across languages. MCP’s open standards allow government IT teams to select models that meet classification and compliance requirements without depending on a single commercial vendor’s roadmap, a form of safe modernization that preserves legacy investment protection while unlocking agentic AI capabilities. All AI actions route through policy-governed orchestration with timestamped, logged, and traceable decision paths, meeting the transparency mandates of the EU AI Act and GDPR’s right to explanation for automated decisions.

IT Service Management: How Does MCP Unify Incident Response?

Large organizations use MCP to unify fragmented operational IT stacks. AI agents autonomously pull incident data from logging systems, cybersecurity platforms, and issue trackers to triage incidents, assign severity levels, and draft root-cause analyses without human intervention. This reduces mean time to resolution and enables IT support contact centers to handle higher volumes of complex tickets while escalating appropriately to human specialists for judgment-dependent decisions.

MCP’s extraordinary momentum comes with real risks that enterprise leaders cannot afford to underestimate. The protocol provides an elegant mechanism for connection, but it lacks built-in frameworks for security policy, authorization, or auditability. Enterprise adoption of agentic AI through MCP is currently outpacing the maturity of the governance controls required to manage it safely. For AI for large enterprises, a transparent security posture and AI governance framework are prerequisites, not afterthoughts. Responsible AI deployment requires governed expansion where every new capability is matched with corresponding controls for bias detection, data loss prevention (DLP), and AI-driven compliance.

Vulnerable Reference Implementations. In early 2026, security researchers identified 30 critical CVEs within 60 days, primarily related to path-traversal and argument-injection flaws in widely copied MCP reference servers. Organizations that adopted these servers without rigorous security review inadvertently imported classic OWASP Top 10 vulnerabilities into their networks.

Identity Dark Matter. AI agents using MCP often bypass traditional Identity and Access Management protocols. Because agents are programmed to find the path of least resistance to complete a task, they frequently exploit stale service identities, local accounts, or long-lived API keys, creating an uncontrolled expansion of privileged access invisible to standard IT governance audits.

Supply Chain Threats. Reliance on third-party, community-built MCP servers introduces supply chain vulnerabilities. An enterprise using an unvetted MCP server is essentially granting a black-box application direct, bidirectional access to its LLMs and proprietary data.

Data Privacy and Sovereignty Risks. When an MCP client reads sensitive data, that information is passed directly into the LLM’s context window. Without strict zero-retention agreements and encryption, regulated PII, HIPAA data, or GDPR-protected information can be inadvertently exposed.

The raw MCP protocol does not inherently mandate Role-Based Access Control (RBAC), Attribute-Based Access Control (ABAC), or audit logging. Enterprise platforms must be secure by design. MCP gateways and governed data platforms are not optional for production deployments. They are absolute requirements for controlled scaling of agentic AI across the enterprise.

While every major CX platform is racing to announce MCP support, the critical differentiator is not whether a vendor supports the protocol. It is how they implement it, what governance they bring, and whether their architecture truly empowers the enterprise to control its own AI future. Avaya Infinity was designed from the ground up to address these requirements, backed by Avaya’s enterprise security heritage and proven reliability serving the world’s largest organizations and government agencies.

Native MCP, Not a Bolt-On. Avaya Infinity integrates MCP directly into its core orchestration engine for native AI orchestration that is not a surface-level integration or third-party add-on. MCP is embedded into the platform’s architecture, enabling true bidirectional orchestration across the entire AI ecosystem with enterprise-grade uptime and automatic failover that ensures continuous service even during peak demand.

Complete Model Agnosticism. Avaya explicitly rejects AI vendor lock-in, delivering true multi-AI provider support. Enterprises using Infinity are entirely free to use Google Gemini, Anthropic Claude, OpenAI models, custom AI models, or specialized open-source models as the cognitive engine for their customer experience. They can swap these models as the AI market evolves without ever breaking their underlying MCP-standardized data integrations, making Avaya the platform for a multi-vendor AI strategy that evolves with the market.

Dual-Role Architecture. Avaya Infinity operates as both an MCP server and an MCP client. As a server, it exposes services like agent status, call summaries, and routing logic to external AI systems. As a client, it consumes services from external MCP-compliant systems.

Enterprise-Grade Security Through Databricks. Avaya’s strategic partnership with Databricks directly addresses the governance gap with cloud-native security that builds on Avaya’s proven compliance track record across regulated industries. By utilizing Databricks to manage the underlying data architecture and serve as the secure data lake, Avaya delivers fine-grained access control through Unity Catalog, strict tenant-aware data segregation for multi-customer environments, immutable audit logging for all AI interactions, advanced threat detection across data pipelines, and seamless integration across both structured and unstructured data sources. This architecture ensures robust data protection and operational integrity across every deployment model.

“Together, Databricks and Avaya empower enterprises to harness domain-specific AI without compromising agility or compliance.”
— Heather Akuiyibo, VP of GTM Integration, Databricks

How Does Avaya Infinity Support Hybrid Cloud and Legacy Modernization?

Unlike cloud-only platforms that force a rip-and-replace approach, Avaya Infinity delivers a hybrid cloud advantage that enables phased cloud migration at the enterprise’s own pace. Organizations can run MCP-connected AI workflows across cloud, on-premises, and hybrid deployments simultaneously, preserving legacy investment protection while progressively modernizing their contact center infrastructure. This hybrid architecture flexibility means enterprises do not need to choose between innovation and stability. They can pursue legacy modernization and strategic cloud adoption as a controlled transition with minimal disruption to ongoing operations.

Avaya’s architecture provides enterprise-grade continuity across every deployment model, with global redundancy, multi-zone architecture, and automatic failover engineered into the platform. For organizations where business continuity is not optional, including healthcare systems, financial institutions, and government agencies, this mission-critical continuity ensures that AI-powered customer interactions remain available and responsive at all times. The enterprise-grade resilience that Avaya has delivered for decades now extends to the AI orchestration layer, providing proven reliability in the environments where failure is not an option.

Tandem Care is Avaya’s model of human-AI collaboration in the contact center, first articulated by Avaya CEO Patrick Dennis. Rather than using AI solely for customer deflection, Tandem Care envisions AI and human agents working in a harmonious cycle. AI agents handle intake, summarization, data retrieval, context assembly, and real-time insights. Human agents manage empathy, escalation, judgment, and relationship-building. The result is empowered agents who deliver better outcomes through a modern user experience that puts the right information in front of them at the right moment.

MCP enables memory and context to persist across both AI and human roles, ensuring continuity. Rather than being sidelined by automation, agents become strategic participants in an AI-augmented workflow. The goal is not fewer humans. It is more empowered humans, backed by AI that understands them, supports them, and never replaces them.

In practice, the Tandem Care model delivers a modern AI experience built on an intuitive agent desktop that consolidates all customer context into a single pane. Before a human agent even says hello, the platform’s AI has already orchestrated real-time data retrieval through MCP, connected to backend systems, pulled the customer’s relevant history, and surfaced tailored recommendations. The agent sees everything at a glance and can greet the customer with full context rather than asking them to repeat information for the third time. This approach creates streamlined workflows that deliver measurable operational efficiency without sacrificing the human connection that builds lasting customer loyalty.

Avaya was the first large enterprise contact center vendor to formally commit to MCP when it announced Infinity platform support in July 2025. CEO Patrick Dennis described it as an “MCP moonshot” that would become core to the product roadmap. But Avaya did not stop at adoption. It joined the foundation shaping the standard’s future.

As a member of the Agentic AI Foundation, the Linux Foundation body providing vendor-neutral governance for MCP, Avaya engineers participate in working groups, community events, and committee nominations tied to the protocol’s development. This membership means Avaya brings customer experience requirements, including governance, security, and compliance, into the discussions shaping how the open standard evolves.

Because Avaya aligns with the AAIF, enterprises are permanently shielded from “protocol drift” or forced migrations. The underlying standards evolve through broad industry consensus rather than unilateral vendor decisions, guaranteeing that AI foundations will remain universally compatible as new breakthroughs emerge. This positions Avaya Infinity as a platform for native AI innovation where new integrated AI capabilities can be adopted as they mature, without waiting for a single vendor’s release cycle. The AAIF now includes nearly 150 members including JPMorgan Chase, American Express, ServiceNow, Autodesk, Red Hat, Cisco, IBM, and Oracle.

Prioritize Open Orchestration Over Proprietary Lock-In. The era of walled-garden AI ecosystems is ending. Enterprise procurement should demand platforms that support open standards like MCP, enabling the freedom to swap AI models and integrate new tools without re-architecting backend systems. Organizations that lock themselves into a single vendor’s AI stack today will face significant switching costs as the market matures.

Invest in Governance Before Scaling. The security risks associated with ungoverned MCP deployments are substantial and well-documented. Before scaling any MCP initiative, ensure that robust access controls, audit logging, and data governance frameworks are in place. This means partnering with platforms that integrate enterprise-grade governance natively, not as an afterthought.

Adopt a Tandem Care Mindset. The most effective contact center AI strategies are not about eliminating human agents. They are about amplifying human capabilities with real-time AI support. Organizations that focus solely on deflection metrics will see short-term savings but long-term erosion of customer loyalty.

Start with High-Impact, Governed Use Cases. Begin MCP implementation in regulated, high-value interaction scenarios such as healthcare, financial services, or complex technical support. These environments demand the governance rigor that separates production-grade deployments from experimental pilots, and they deliver the clearest long-term ROI and value justification for broader rollout.

Evaluate Platforms on Security Posture, Not Just Feature Lists. Every CCaaS vendor is announcing MCP support. The critical evaluation criteria should include how data governance is handled, which partners secure the data layer, whether the platform supports fine-grained RBAC and audit logging, and whether the vendor offers transparent pricing with predictable costs rather than opaque consumption-based models that escalate over time.

Plan for Safe Modernization, Not Forced Migration. The best MCP platforms support enterprise modernization as a phased journey, not a cliff. Look for vendors that offer hybrid architecture flexibility, legacy investment protection, and a clear path from on-premises to cloud that preserves your existing workflows, data, and integrations while progressively adding agentic AI capabilities through controlled scaling.

The critical test question: If a breakthrough model is released next month, how fast can you use it without re-engineering your workflows?

White Papers

• The Model Context Protocol: A Status Report for Enterprise CX Leaders (white paper)
• The Importance of Being Open for AI (white paper)

Solutions and Platform

• Secure MCP Integration with Avaya (solution page)
• Avaya Infinity Platform (product page)
• Avaya MCP FAQ (FAQ page)

Blog Posts

• Why Avaya Is Investing in Open AI Foundations (blog)
• Avaya Announces Native MCP for Infinity (press release)

Related Insights Pages

• What is CCaaS in 2026? (insight)
• What is UCaaS in 2026? (insight)
• Agentic AI for Contact Centers in 2026 (insight)

• Agentic AI Foundation (AAIF) Vendor-neutral governance for MCP under the Linux Foundation
• Linux Foundation: AAIF Formation Announcement (December 2025)
• Anthropic: Model Context Protocol Original announcement and open-source release
• MCP Official Specification Protocol documentation and technical reference
• Gartner: Top Strategic Technology Trends AI connectivity and enterprise architecture research
• OWASP Top 10 Web application security risks framework
• Databricks Enterprise data governance platform, Avaya MCP security partner
• MIT Sloan School of Management Research on open-weight vs. proprietary AI model economics