In the telecommunications industry, access to advanced network capabilities is key to delivering higher quality and more efficient services to customers - from resource management and fraud prevention, to improved connectivity and device monitoring.
In the coming years, many of these digital interactions will be managed by intelligent agents operating through MCPs: AI systems capable of reasoning, deciding which tools to use and executing actions autonomously and in real time.
The Model Context Protocol (MCP) is an open standard, launched by Anthropic in November 2024, that allows AI assistants to connect with external data sources and tools via a common protocol, eliminating fragmented integrations.
It functions as a "USB-C for AI applications": it standardises how models read, write and execute actions on data, APIs and external systems. In the telco sector, this opens the door to a new generation of agent-driven API-driven services, where the intelligence lies not only in the network, but also in how these agents decide, orchestrate and monitor its use.
The evolution of network APIs towards MCPs not only improves technical efficiency, but redefines the way digital capabilities are built, consumed and scale in an increasingly automated and distributed environment.
What are intelligent agents and why do they need MCPs?
Intelligent agents are language model-based systems (LLMs) capable of planning, reasoning and executing autonomous actions on different digital resources. Rather than simply answering questions, these agents can access data, interact with tools and adapt in real time to the environment and the conversation.
However, their true potential is unlocked with the Model Context Protocol (MCP). MCP is an open protocol that:
How does MCP work?
The Model Context Protocol (MCP) establishes a standardised framework for intelligent agents to interact with external systems in a contextual, secure and reusable way. Rather than relying on ad hoc integrations, MCP defines how agents discover, execute and manage tools, data and APIs through a common layer of interoperability.
Its basic architecture consists of the following elements:
For example, a banking application that integrates a risk engine with an agent as a risk co-pilot can communicate with a CCM to see what tools it has available, and explore one of them to see if a mobile line shows signs of risk (such as a recent SIM change).
This access is standardised and traceable, without the need to develop a specific integration for each network API. MCP allows the agent to dynamically "discover" available tools and invoke them according to their logical flow. The logic or flow that could follow would be as follows:
MCP therefore represents a structural shift in intelligent agent architecture from fragmented, ad hoc models to a modular, scalable and contextualised approach. In the telco sector, this evolution not only reduces technical friction, but also enables a new generation of intelligent services, capable of dynamically adapting to the environment, integrating with multiple network APIs and delivering greater operational and business value.
Advanced use cases: CCM and AI agents in action
On-demand connectivity for high quality video: An intelligent agent managing video conferencing sessions needs to ensure optimal network conditions for each user. Without MCP, every integration with network APIs - such as QoD - must be developed and maintained manually.
With MCP, the agent can discover and access these APIs in a standardised way, without the need to implement specific connectors. MCP does not orchestrate the network directly, but allows the agent to maintain context about the session, select the appropriate API (e.g. QoD request) and record the result for future decisions.
Fraud prevention with multiple risk signals: A risk officer tasked with detecting fraud needs access to signals from different sources - such as usage patterns, device history or recent mobile line changes - to make informed decisions.
Without MCP, each source must be integrated separately, which introduces complexity, context fragmentation and maintenance difficulties.
With MCP, the agent can access these signals through a shared, standardised context space, where each tool exposes its information in a consistent way. This does not mean that MCP merges data or makes decisions on its own, but rather that it helps the risk agent to consult, update and reason on distributed contextual information, maintaining traceability and consistency throughout the flow.
A new frontier for carriers and customers
The Model Context Protocol (MCP) not only improves the way intelligent agents consume APIs, but also opens a strategic avenue for telecom carriers to expand their role in the digital economy.
By adopting MCP, carriers can evolve from simple connectivity providers to enablers of intelligent services. This enables capabilities such as adaptive connectivity, identity validation, fraud prevention or on-demand quality of service to be offered in a more modular, interoperable and autonomous agent-ready way.
Far from replacing existing APIs, MCP acts as a standardised interaction layer, where agents can discover, access and use network tools - such as those enabled by Open Gateway - without relying on custom integrations or embedded logic.
This approach favours the exploration of new use cases with less technical friction, while preserving control over access to and use of sensitive information exposed through its APIs.
