Generative AI has quickly gone from novelty to utility. They can draft emails, summarise reports, analyse data, write code, create media, and respond fluently to complex queries. But there's a major catch: they're still disconnected from your business context, their own data is often somewhat out of data, and they tend to make things up.

That’s not to say context hasn’t improved. It has. Models are no longer sealed off from the world. They can search the internet for information and process uploaded files. But these solutions are still manual and static. If you turn to Claude for help with refining your content strategy, it can't check which topics generated the most traffic last quarter unless you feed it the data. If you want help to analyse trends, it can't pull your latest customer segmentation or engagement stats on its own.

Model Context Protocols (MCPs) aim to solve this. They connect AI directly to your business systems—your CRMs, analytics tools, knowledge bases, and project files—so it can access the information your company already has. The result: AI that’s not just clever, but context-aware and genuinely useful.

What Are Model Context Protocols?

Model Context Protocols (MCPs) are an open standard introduced by Anthropic in late 2024. They act as standardised gateways, defining how AI models—such as Claude—can connect to and work with external systems.

MCPs give AI models access to:

Resources – Information the AI can read for context, such as your documents, a CRM record, a knowledge base article, analytics reports, or team communications.
Tools – Actions the AI can take, like searching your database, emailing, calling an external API, or triggering an automated workflow.

Think of them as a universal connector for AI. Just like USB-C lets you plug various storage devices and tools into your computer, MCPs let any compliant AI model request a Tool or Resource.

MCPs use a simple, lightweight client-server structure to connect each data source or service. They are open-source, model-agnostic, and already supported by hundreds of community-built connectors.

MCPs are supported by a growing ecosystem of software vendors, SaaS providers, and open-source contributors. For instance, Slack might release an MCP server for its messaging platform, while HubSpot could provide one for its CRM. Internal development teams can also build custom MCP servers to connect proprietary or unique internal systems. This flexibility means that both off-the-shelf and tailored integrations are possible, depending on the complexity of your systems and the level of control required.

Getting Started with Model Context Protocol (MCP)

The best way to understand Model Context Protocol (MCP) is to try it out. For now, the easiest way to do it is with the Claude Desktop App. This is the downloadable version of the Claude AI Chatbot. Once you have it, you can add MCPs that plan tasks better, access Google Maps, or even organise and rename some files on your computer.

For now, there are few fairly technical steps needed to get MCP working on your computer. You might have to add some extra software and you will have to edit a configuration file. These steps look more difficult than they really are, so don’t be discouraged. Editing a configuration file looks daunting but it’s hard to go really wrong. It either works or it will need editing a again until it does. You won’t break anything if you make a mistake. You might find this guide to getting started with MCPs helpful. We’re very confident that adding MCPs will become much more user-friendly soon.

Anthropic, who introduced MCP, created some example MCP servers that you can use. Subsequently other companies like HubSpot, Make and Stripe have added their own. You can find out more about what’s available by looking at Anthropic’s launch documentation.

Once experience has given you a clearer idea of MCPs can add superpowers to your AI tools, it becomes easier to imagine what MCP servers your own business might want to develop for itself or your customers. If your company already has an API that people inside or outside your organisation can use, you might want to consider whether offering it as an MCP will empower users even more.

Why MCPs Are Useful for Businesses

For businesses, MCPs are arriving at just the right moment. Many organisations have experimented with AI in siloed ways – a chatbot here, an analytics helper there – only to find these AI tools lack real business context and are hard to connect with enterprise data.

By offering a universal connector, MCPs dramatically cut the cost and complexity of integrating AI with the systems that matter. Instead of spending months building custom adapters for each model–tool combination, companies can adopt the standard and get a plug-and-play solution.

This has several concrete benefits:

Richer, more accurate insights: MCPs enable AI to access live, organisation-specific data, making outputs more relevant and up-to-date. For example, an AI sales assistant could automatically incorporate today’s customer engagement metrics or inventory levels into its recommendations.
Time savings and scalability: MCPs streamline integration across the tech stack, reducing project timelines and making it easier to scale AI solutions across cross departments, functions, or client accounts.
Flexibility and Futureproofing: Interoperability is good news for business leaders seeking agility. Open protocols reduce vendor lock-in, making it easier to swap out AI providers or upgrade technology. By decoupling integrations from specific vendors or models, you can integrate AI capabilities into your workflow or product offerings without worrying that you’ll be left stranded on a technological island.
Secure, governed integration: MCP was designed with enterprise needs in mind. Connections can be configured to respect permissioning, and data stays within your controlled environment when required. By using two-way authenticated links (clients and servers) rather than opening up broad internet access, MCP lets AI reach into internal systems in a secure, auditable way.

Challenges and Considerations

While MCPs offer significant advantages, they also present challenges worth considering:

Technical and Adoption Hurdles As a newer standard, MCPs currently require technical expertise to implement effectively. Documentation and best practices are still developing, which can make troubleshooting more complex.

Security and Governance Any technology connecting AI to business systems requires careful security consideration. With MCPs, organisations must establish authentication, define permission boundaries, and implement monitoring to prevent unauthorised data access. Proper configuration is essential to avoid risks like over-privileged access or prompt injection attacks where malicious inputs manipulate AI behaviour.

Ecosystem Maturity The MCP ecosystem, while growing rapidly, is still developing. Some specialised systems may not yet have pre-built connectors available, potentially requiring custom development. As with any emerging standard, early adopters should be prepared to stay on top of updates and implement refinements.

Are MCPs really interoperable?

MCP might have begun as an Anthropic initiative, but a broad coalition is emerging to make MCP the de facto way AI connects to tools.

Other major AI companies, including OpenAI and Google DeepMind, have indicated they will add MCP support to their models and tools. Enterprise tech companies like AWS and GitHub are also on board. AWS has been involved in the community, ensuring AWS data services can interface via MCP. GitHub has adopted MCP within its own Copilot ecosystem, allowing code-focused AI to fetch context from project data.

In short, it looks promising for interoperability. Competing standards may exist, but MCP’s rapid adoption suggests a convergence. We are essentially seeing the birth of a cross-platform standard – much like how all modern web browsers adhere to HTTP/HTTPS protocols, so any browser can talk to any website or a common data standard, like SQL or REST APIs, that let you swap out components freely.

While it’s still early days and each vendor will have its own rollout timing, the commitment to a common protocol signals that MCP is likely to remain interoperable across the AI industry. The focus now is on strengthening the standard collaboratively–for example to add new capabilities or improving security–rather than fragmenting into closed ecosystems.

How You Can Use MCPs in B2B Marketing and Professional Services

How can these abstract benefits translate into day-to-day business scenarios? In both marketing and professional services, MCPs unlock new possibilities for AI assistance by fusing siloed information and automating complex workflows. Below are a few illustrative use cases.

Unified Customer Insight

Imagine your marketing team preparing a personalised campaign for a major client. An AI assistant equipped with MCP could simultaneously pull the client’s latest product usage stats from an internal database, recent support tickets from the CRM system, and relevant case studies from the content repository. It could then draft a tailored proposal or campaign strategy that weaves in all these up-to-date insights.

Without MCP, gathering that information would require manually exporting or expensive custom integrations to get the data out of each system. With MCP, the AI can query each source on-the-fly and deliver a cohesive output. This saves time and grounds marketing content in the very latest data.

The same approach can power sales enablement tools – for example, an AI sales coach that can listen to live sales calls and retrieve product details, pricing history, and relevant marketing materials in real time to assist the representative.

Client Service and Research

Maybe you’re a consulting firm that needs to answer complex client questions by drawing on your company’s diverse knowledge. With MCP, an AI can act as a research concierge. When a consultant poses a query, the agent might fetch the latest industry regulations from an external database, pull internal precedent documents or project reports, and even run calculations via an API – all in one conversational flow.

Crucially, MCP maintains context across these tools, so the AI can combine the information before responding. For instance, in financial advisory, an AI agent could retrieve a client’s portfolio data, analyse real-time market feeds, and then produce recommendations compliant with the client’s risk profile and regulations. In essence, MCP-enabled agents can serve as junior analysts that instantly tap the firm’s knowledge base and external data, boosting the capacity of human experts.

Enhanced Customer Support Bots

Many B2B companies offer complex products or services that generate a lot of support queries. An MCP-powered support chatbot can break through the old limitations of FAQ bots. For example, a cloud software company could deploy a bot that, via MCP, has access to the product documentation, the user’s account data, and even the system’s status monitoring API. If a customer asks, “Why am I getting this error?”, the AI can check the knowledge base for that error message, look up the customer’s configuration from the account database, and maybe run a diagnostic through an API. It can then formulate a single, informed response, offering a precise answer that a human agent would otherwise have to compile manually.

Multi-Agent Teamwork

MCP makes it practical to use multiple specialised AI agents in tandem, which is a growing trend in enterprise AI. For instance, a marketing department might have one AI agent focused on lead qualification (scanning inbound inquiries or website chats and assessing prospects), another agent specialised in content generation (writing blog posts, social media content, etc.), and perhaps a third handling analytics reporting.

With MCP acting as the integration layer, these agents can share context and orchestrate tasks smoothly. One agent can hand off to another along a defined workflow – say, the lead-qualifier flags a high-potential lead, then triggers the content generator agent to draft a personalised follow-up email for that prospect. All the while, they might use common resources (e.g. a shared CRM data source) via MCP, ensuring consistency.

Such coordinated AI teams were very hard to implement before, because of each agent needing its own integrations. MCP provides a common highway for agent-to-agent and agent-to-data communication, making these complex workflows workable and robust.

Implementing and Publishing MCPs in Your Organisation

One strength of MCP being an open standard is that any company can create and share MCP integrations for their own systems or data.

Implementing MCP in your organisation involves two steps:

Deploying MCP servers for the data sources or services you want to expose,
Using an MCP-compatible AI client (or platform) to consume those servers.

Fortunately, the ecosystem is maturing rapidly to support this. Anthropic’s open-source repository already offers pre-built MCP server connectors for many popular enterprise platforms – from Google Drive and Slack to GitHub and SQL databases. This means you might download or adapt an existing connector rather than building from scratch.

If a needed connector isn’t available yet, companies can develop their own MCP server using the open specification and SDKs (software development kits) provided in multiple languages. Many organisations have done so already – the community has created MCP servers covering everything from proprietary databases to niche SaaS tools.

Implementing a server typically involves wrapping your system’s API or database queries in the MCP format. This defines the Resources and Tools available to an AI.

It can be as private or as public as you need: internally to break down your organisation’s data silos for your AI systems, and externally to offer new intelligent services or integrations to clients.

In-house, this might mean your engineering team builds an MCP interface to break down your organisation’s silos so your company’s AI applications can start leveraging that data.

Externally, it opens up new productisation opportunities. Just as companies once exposed REST APIs to let others access their services programmatically, now they can expose MCP endpoints to let AI agents tap into those services in a controlled way. You could publish an MCP connector for your platform as a value-add. AI-ready integration could become a selling point, allowing your product or data to be easily used by the growing field of AI-powered applications across vendors.

Over time, we may see entire marketplaces of MCP integrations much like app stores or API marketplaces – a development that could lower the barrier for companies to deploy rich AI solutions without re-inventing the integration wheel for each new project.

A New Integration Layer for the AI Era

Most of the noise around generative AI has focused on model capabilities—bigger, faster, more human-like. But for businesses, the more urgent question isn’t what AI could do. It’s what it can’t do without the right context.

MCPs are quietly shifting focus from the model itself to the infrastructure that surrounds it. They won’t solve everything—but they address one of the biggest blockers so far: making AI actually understand and work with your business. They make AI part of the business stack—able to access live data, take meaningful actions, and coordinate across systems.

This is where the potential lies. The businesses that extract the most value from AI won’t be those chasing the latest model. They’ll be the ones investing in integration, building the connective tissue that lets AI see, understand, and act within their specific operational context. They’ll be the ones wiring intelligence into the flow of real work.

The question now isn’t just what can AI do? It’s whether your systems and culture are ready to let it do it. If your business wants to move from AI curiosity to AI capability, this is where your focus should be.

AI is only as effective as the strategy behind it.

At 1827 Marketing, we help B2B marketers make smart decisions about when and how to use emerging tools. If you’re exploring how to use AI more meaningfully in your marketing, we’d love to talk.

Get in Touch

Connecting AI to Business Reality: Why Model Context Protocols (MCPs) Matter for B2B