- MCP is a standardized way for applications to provide context to LLMs
- It was first introduced by Anthropic in November 2024
- Will telco vendors also try to get into the game? Analyst Dean Bubley thinks that it may already be too late
What’s an agentic AI-related trend that’s coming into focus for telcos and a gaggle of other business sectors? It's model context protocol (MCP), an open protocol that Anthropic introduced in 2024. Few telecom equipment vendors are currently building with the protocol, and Dean Bubley, founder of Disruptive Analysis, said it may be too late for them to benefit.
MCP "provides a standardized way to connect AI models to different data sources and tools. MCP enables you [to] build agents and complex workflows on top of LLMs and connects your models with the world," according to the MCP Project website.
It allows systems to provide context (aka data) to AI models in a manner that’s generalizable across integrations, said engineer Yoko Li in an Andreessen Horowitz article about the technology. “The protocol defines how the AI model can call external tools, fetch data, and interact with services,” she noted.
“MCP in 2025 is kind of like HTTP in the early 1990s — it has the potential to change how people interact with businesses and services, and create new types of businesses altogether,” Cloudflare VP of Product Rita Kozlov told Fierce in April this year.
In a way, we can compare how MCP interoperates with an AI network with the way that 5G APIs work with a 5G network. Both are designed to connect aspects of the technology with other elements.
What's the telecom vendor angle?
So, will telecom vendors get into the MCP game?
“I can certainly imagine that the major vendors will start to work with (or implement their own) MCP – for instance as a gateway to their APIs,” said Bubley in an email. “I’d imagine that Nokia and Ericsson/Vonage/Aduna have at least got some people tinkering in the labs on this.”
“I suspect the initial uses will be internal for their own developer and product teams, where they’re using AI agents already,” Bubley added. “So probably something around CRM/chatbots giving them access to specific resources, or inside the OSS/BSS, especially if TM Forum adds MCP into its ODA architecture. I’d also imagine we’ll get some press releases about putting an MCP pathway into CAMARA/Open Gateway APIs as a trial, but don’t imagine it will be commercialized any time soon.”
Bubley stated that he had already seen some proposals for MCP-telecom extensions. He added that he expects that mobile telecoms will make a lot of noise about MCP servers for telecom uses.
“I think it’s too late for telcos to shape MCP overall – that ship has sailed, and is accelerating quickly, with or without telcos as a specific vertical (among many),” Bubley said.
Fierce Network's take
Today, MCP is being driven by AI labs, cloud hyperscalers and platform companies like OpenAI, Anthropic, Microsoft and Google — not telcos because it's fundamentally a software coordination layer, not something telcos have historically influenced.
However, MCP could become as fundamental to AI workloads as TCP/IP is to data networks. If telcos miss this layer, they will have no say in how AI systems access network intelligence such as location, latency, QoS. They risk being just “dumb pipes” again — this time for AI coordination traffic.
Telcos now rent compute from AWS, Azure, among others. If they don’t engage with MCP development, they’ll depend on hyperscalers for AI integration, further eroding control over service layers, and may struggle to differentiate their networks for AI-native applications.
Operators could contribute network context APIs that feed MCP models real-time radio or routing info, and edge orchestration logic to help LLMs decide where to run tasks. But if they’re not at the table, those features won’t be built in — and networks will stay invisible to AI agents.
If this sounds a bit familiar, you're right. We've already seen something similar happen with 5G slicing and the cloud-native core — telcos hoped to monetize it, but the value went to cloud players. It's also similar to what happened with WebRTC — telcos didn’t engage early, so over-the-top players won the market.
To remain relevant, telcos should propose open MCP extensions for network context, offer telecom toolchains such as service provisioning or mobility to AI agent frameworks, or join or fund open-source MCP projects.
Stay tuned to see if any of that happens. We'll be watching.