Is MCP Likely to Achieve Industry Adoption? (hint: we think so)

Picture this: you’re building an AI agent that needs to talk to your CRM, pull data from your database, and send notifications through Slack. Right now, you’d have to build custom integrations for each tool, dealing with different APIs, authentication methods, and data formats. It’s a mess.

That’s exactly the problem the Model Context Protocol (MCP) is trying to solve. Think of it as USB-C for AI agents—one standard way for AI models to connect with any tool or data source.

The big players are already on board

Here’s what caught my attention: three of the four major AI labs have already lined up behind MCP. We’re talking about OpenAI, Google, and Anthropic—plus Microsoft is throwing its weight behind it too.

OpenAI has baked MCP into their Agents SDK, ChatGPT desktop app, and their new Responses API. Google DeepMind integrated it into the Gemini SDK. Microsoft went all-in with native Windows support and their AI Foundry platform.

The only holdout? Meta. They’re still focused on open-source frameworks like LlamaIndex and LangChain, but no MCP integration yet.

That vendor line-up is exactly why the directory of MCP servers available to marketing teams keeps growing every month. If you’re deciding whether a given integration needs the full MCP treatment or just a direct API call, see MCP vs API for the trade-offs.

Why this matters now

Current AI function-calling APIs are a nightmare. Every vendor does it differently. If you want to build an agent that works across platforms, you’re basically starting from scratch each time.

MCP solves this the same way REST solved the complexity of SOAP in the early 2000s. It gives developers a standard way for AI agents to discover tools, request resources, and manage context across conversations.

The timing is perfect too. The agent platform ecosystem is still forming, and standards set now will have massive network effects later. Plus, enterprise buyers are demanding vendor-neutral solutions to reduce AI deployment risks.

How fast is MCP growing?

The adoption numbers were impressive from day one, and they’ve only grown since. In its first 9 months, MCP already had thousands of integrations; by Q1 2026, that had scaled to over 17,000 MCP servers, and by July 2026 Anthropic reported SDK downloads had passed 400 million a month, a 4x increase in a year (more in is MCP dead?). To put the early speed in perspective, here’s how MCP’s launch compared to other protocol adoptions:

That’s unprecedented speed for a new protocol, and the pace has held up: MCP has kept adding vendors and servers every quarter since.

The missing pieces

MCP isn’t perfect yet. There are some real challenges:

Security is the big one. AI agents with tool access create new attack vectors—token theft, prompt injection, tool poisoning. The major providers are being cautious, rolling out gated previews with strict guidelines.

Governance is another issue. Right now, it’s still vendor-led by Anthropic. For true industry adoption, MCP needs neutral governance—think how GraphQL moved to the GraphQL Foundation in year three.

Competition could fragment things. Proprietary SDKs, gRPC alternatives, or other standards could split the market.

We go deeper on these risks, and how to plan around them, in MCP’s key challenges for marketers.

Will MCP actually win?

I’m putting the odds at 65% that MCP becomes the dominant standard within five years. Here’s my reasoning:

The vendor alignment is incredibly strong (8/10). Having Anthropic, OpenAI, Google, and Microsoft all backing the same protocol is rare.

Developer experience is solid (7/10). It’s built on familiar JSON-RPC patterns, has good documentation, and growing open-source server ecosystem.

But security concerns (5/10) and lack of neutral governance (4/10) are real drags on adoption.

Revisiting the forecast: how has the 65% odds held up?

That forecast is now testable rather than theoretical. Here’s how each input has moved since July 2025:

Vendor alignment and governance, the two weakest inputs, have both moved decisively in MCP’s favour. Security is the one score that hasn’t fully resolved, though it moved in July 2026 when the MCP 2026-07-28 spec aligned authorisation with production OAuth 2.0 and OIDC, so servers connect to identity systems like Entra or Okta without workarounds. And as we cover in is MCP dead?, MCP’s role is narrowing toward the multi-user, team and enterprise workflows where that trade-off makes the most sense, exactly the audience this article was written for. On balance, the original 65% looks conservative rather than wrong.

What we’re doing about it

At Growth Method, we think MCP has the momentum, money, and product-market fit to win. The security issues are being addressed, and neutral governance has followed, with the Linux Foundation now stewarding the protocol, just as we expected.

So we’re standardising our data integrations on MCP, with Zapier and custom integrations as fallbacks. It’s a bet on where we think the ecosystem is heading.

The bottom line? MCP feels like being at the REST vs SOAP crossroads in 2002. The simpler, more practical solution usually wins—and that’s exactly what MCP represents for the AI agent world.

If you’re building anything that connects AI to external tools or data, MCP is worth learning now. The network effects are already starting to kick in, and being early to a winning standard pays dividends for years.

About Growth Method

MCP standardises how AI models talk to your tools. Growth Method is the agentic marketing platform for B2B teams built on top of that standard: we bet on MCP early by standardising our own data integrations on it, and that bet compounds every time a new MCP server ships, one connection, not a fresh integration for every tool in a fast-moving stack. Plan, launch and analyse campaigns from your live marketing data, with your AI agents pulling straight from the tools you already run.

We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.

Laura Perrott, Colt Technology Services

Get started to connect your stack and put MCP’s standard to work.

Frequently asked questions

Is MCP still gaining ground in 2026?

Yes. The ecosystem counts more than 17,000 MCP servers (Q1 2026) and over 400 million monthly SDK downloads (July 2026, a 4x increase in a year), and the protocol is now stewarded by the Linux Foundation as an open standard. Anthropic, OpenAI, Google, Microsoft and AWS have all shipped deep MCP support, exactly the vendor alignment this article predicted back in July 2025.

Has Meta adopted MCP yet?

Not publicly. Meta remains the notable holdout among the major AI labs, still focused on open-source frameworks like LlamaIndex and LangChain rather than adding native MCP support.

What’s the biggest risk to MCP’s dominance?

Governance, once the weakest input in our forecast, has largely resolved now that the Linux Foundation stewards the protocol. The live risk today is different: a vocal developer backlash argues MCP’s token overhead isn’t worth it for solo, CLI-first workflows. That argument is strongest for single-user use and weakest for the multi-user, team environments MCP was built for.