Model Context Protocol
What it is and why it matters
Model Context Protocol (MCP) is an open standard that defines how AI applications connect to external tools, data sources and systems through a consistent interface. Often described as a USB-C port for AI, MCP provides applications with a simple way to connect to databases, business systems and other enterprise tools. This helps organizations reduce complex integration work, improve efficiency and unlock real-time insights, making AI agents more useful in everyday business scenarios.
The emergence of Model Context Protocol
MCP grew out of a practical challenge in AI development: Teams often had to build separate, custom integrations for each tool or platform, making agentic AI systems harder to maintain and scale. In November 2024, Anthropic introduced the open source MCP to standardize how AI agents connect to external data sources and tools.
The protocol quickly gained traction because it addressed a clear need to make AI integrations easier to build and manage across systems. Within a few months of launch, contributors from Anthropic, Microsoft, GitHub, OpenAI and the broader developer community began extending MCP, including improvements on security, tool discovery and enterprise readiness.
In December 2025, Anthropic announced that it was donating Model Context Protocol to the Linux Foundation’s Agentic AI Foundation. This transition marked a significant milestone. It signaled to businesses and developers that the standard is now backed by a trusted foundation, supporting compatibility and reliability as organizations build AI-powered tools and workflows.
Why Model Context Protocol matters
When Anthropic introduced the Model Context Protocol, the goal was to break down data silos by giving AI systems a standardized way to securely connect with business applications, data sources and tools.
Before MCP, every AI application required custom integrations and each tool needed its own connector. Security models varied by application, and maintenance costs grew as systems expanded.
With the protocol, a standard interface connects AI applications to tools and data. Organizations can reuse integrations across multiple AI systems. Developers spend less time building connectors, and security and AI governance become more consistent.
By simplifying how AI applications discover and interact with enterprise resources, MCP reduces integration complexity, strengthens interoperability and supports scalable AI-driven workflows. The result is AI that delivers more relevant, personalized responses based on real-time business context.
Model Context Protocol in today’s world
Growing industry adoption. Since its launch in late 2024, MCP has quickly gained popularity across the AI ecosystem. Adoption is growing among AI assistant platforms, enterprise software vendors, cloud providers and open source AI frameworks that want a straightforward way to connect AI systems with tools, applications and data.
Major technology players have already announced support, integrations or contributions to MCP. Anthropic, which introduced the protocol, has been joined by OpenAI, Microsoft, GitHub and others that are incorporating the technology into their AI products, developer tools and cloud services. This growing participation signals strong industry momentum toward establishing a shared standard for how AI systems discover and interact with enterprise resources.
MCP and agentic AI. The Model Context Protocol has become critical as organizations move beyond conversational chatbots toward AI agents that can complete tasks, access information and interact with business applications. The MCP provides AI agents with a reliable way to retrieve data, run tools and work across various enterprise systems. Without a common protocol, companies end up building numerous custom integrations, making agentic systems harder to scale.
Learn more about agents and MCP tools in the links below.
Threat modeling for agentic systems
As MCP and agent-to-agent protocols expand what AI agents can do, security teams need new threat-modeling approaches. This article outlines risks such as goal misalignment, tool-instruction abuse and cascading trust failures, along with practical controls for safer agentic systems.
Reliable data access for AI agents
AI agents need more than powerful models to support real operational decisions. This article explains how Retrieval Augmented Generation (RAG) and MCP work together: RAG retrieves unstructured enterprise knowledge, while MCP enables AI agents to connect to current structured data from operational systems in real time.
Unlocking agentic AI with MCP tools
Integrating Model Context Protocol into SAS Retrieval Agent Manager helps enterprise AI agents retrieve information, reason across systems and act within defined governance frameworks. With structured tool execution, reliable governance and flexible deployment, organizations gain a scalable foundation for automating real-world business processes.
How industries use Model Context Protocol
From financial services and insurance to public sector and health care, MCP gives AI agents a standard way to connect with business systems, retrieve context and act across systems with defined controls.
The Model Context Protocol … addresses one of the biggest challenges in enterprise AI: how to make AI systems interact with live operational environments without sacrificing governance, safety or clarity. Ramya Mopivedi Senior Product Manager SAS IoT Product Management
See MCP in action | Explore MCP with SAS
Model Context Protocol can expand how AI agents and AI models operate across data science workflows. In this webinar, learn how to empower AI agents with SAS Viya MCP Server, which connects them to SAS analytics, decisioning capabilities and integrated tools. Learn about practical use cases for adding advanced analytics to agentic AI systems.
How Model Context Protocol works
MCP is an open standard that allows AI applications to access databases, enterprise systems, development tools and creative tools without needing one-off integrations. This means businesses can use AI agents more easily, reducing setup complexity and supporting more consistent integration across platforms.
The protocol operates on a client-server architecture, in which an AI application connects to one or more MCP servers via dedicated MCP client connections. For each server, the AI app creates a separate client to manage that connection to that server. In this setup, the agent can request information or actions in a structured way, while the server enforces rules about what data and tools are available. In this manner, AI agents can securely access approved systems without compromising governance or security.
By providing a standardized connection layer, the protocol reduces duplicated integration work, accelerates development and supports stronger security controls. It also helps AI agents operate within a relevant business context, whether when retrieving CRM data, scheduling meetings or analyzing enterprise databases.
MCP vs APIs
Application programming interfaces (APIs) are a set of rules that allow software programs to exchange data and perform actions. They define how applications communicate, including requests, responses and security requirements. APIs save software developers time. Since APIs can be reused, developers do not need to build these entirely on their own.
Model Context Protocol, on the other hand, provides a standardized way for AI applications, such as large language models, to discover and interact with tools, data sources and services. Rather than replacing APIs, MCP typically sits on top of them, offering a consistent interface that AI systems can more easily understand and use.
MCP security and governance
While Model Context Protocol provides a standardized way for AI systems to connect with enterprise tools, it is not a security governance or compliance framework. Organizations must continue to be responsible for:
- Identity management (who is allowed to connect).
- Authentication (verifying users/agents).
- Authorization (defining what actions are permitted).
- Data governance (ensuring compliance with privacy and regulatory rules).
- Auditability (maintaining logs and forensic trails of agent actions).
- Human approval workflows (ensuring human oversight for sensitive or critical actions).
MCP architecture
The Model Context Protocol architecture provides a standardized way for AI agents to securely connect with external systems and data sources. Instead of integrating with each application individually, the AI agent communicates through an MCP client and MCP server, which act as a unified gateway to enterprise resources such as databases, CRM and ERP systems, analytics platforms, APIs and code repositories. This modular approach simplifies integrations, improves scalability and enables AI applications to access business context consistently across multiple systems.
- An AI agent is the reasoning engine and interaction point; an AI agent determines exactly which data or tools are needed to complete a task.
- The MCP client lives within the host environment. It translates the agent's intent into the standardized protocol and manages the secure connection.
- An MCP server is the stateless bridge that safely exposes specific capabilities – translating protocol requests into usable queries.
- Enterprise systems are the downstream data layer where the server executes operations against databases, CRMs, ERPs, specialized SAS analytics platforms, APIs or document repositories.
- Databases store structured and unstructured business data, such as customer records, transactions and product information.
- Customer relationship management (CRM) systems manage customer interactions, sales activities and customer support histories.
- Enterprise resource planning (ERP) systems integrate core business functions such as finance, procurement, inventory and human resources.
- Analytics platforms collect and process business intelligence, metrics and reporting data.
- Application programming interfaces (APIs) enable communication between software applications by exposing data and functionality.
- Repositories store code, documents and other version-controlled assets in platforms such as GitHub or GitLab.
Recommended reading
-
Article
Trustworthy AI: How to move from ambition to ROITrustworthy AI is artificial intelligence technology that is designed, developed and deployed with human-centricity and other core principles in mind. Learn what it means to trust AI in the face of constant, real-world business pressures. And see why treating trustworthy AI as a repeatable operating model provides a long-term business advantage – and financial returns.
-
Article
Understanding digital twin technologyDiscover how digital twin technology transforms industries from health care to manufacturing as it shapes the future of data modeling and process improvements. For example, digital twins can simulate conditions and predict outcomes, offering insights into potential improvements. This allows organizations to test and refine their operations digitally, speeding development and reducing costs to help them become more agile and competitive.
-
Article
Containing health care costs To ensure payment integrity, health care organisations must uncover a broad range of fraud, waste and abuse in claims processing. Data-driven analytics – along with rapid evolutions in the use of computer vision, document vision and text analytics – are making it possible.