guide

What Is MCP and Why Does It Matter for Marketing Teams?

The Orbitable Team·AI & GTM·5 May 2026·5 min read

MCP, the Model Context Protocol, is an open standard that lets an AI model like Claude call external tools and data sources through one consistent interface, instead of every application building its own one-off integration. For a marketing team, it matters because it turns whatever software you already use, including a 50-agent platform, into something an AI assistant can operate directly, on request, without you opening a separate app.

What Is MCP, Exactly?

MCP was introduced by Anthropic in November 2024 as an open specification, not a proprietary Anthropic feature. It defines a client-server relationship: an MCP server exposes a set of "tools" (actions it can perform), and any MCP client, Claude included, can discover those tools and call them mid-conversation. The protocol standardises the handshake, so a developer builds one server instead of a separate custom connector for every AI product that might want to use it.

Before MCP, connecting an AI model to a piece of software meant writing bespoke glue code for each pairing. MCP replaces N custom integrations with one standard interface that any compliant client can speak.

Why Marketing Teams Have Mostly Ignored It

MCP launched as developer infrastructure. The early adopters were engineering teams wiring Claude into codebases, databases, and internal tools. Marketing software vendors were slower to build MCP servers because most marketing platforms are designed to be logged into, not called into from outside.

That gap is the reason MCP has stayed invisible to marketing teams even as it became standard practice in engineering. The protocol was ready. The marketing tooling wasn't built for it yet.

How Orbitable's MCP Server Works

Orbitable runs 50 specialist marketing agents across 10 squads, coordinated by the Dispatcher, which orchestrates which agents run and in what order. Orbitable built an MCP server that exposes every one of those 50 agents as a callable tool to Claude, and to any other MCP client.

That means the fleet is no longer confined to Orbitable's own interface. A user connected via MCP can ask Claude to run a specific agent's task, and the request goes to the same agent, working from the same shared world model, that would run if triggered inside the app. Every agent draws on one world model per customer, meaning ICP, brand voice, product facts, competitors, and the uploaded knowledge base, so an agent called through Claude has the same context as one called from inside Orbitable. Nothing gets re-explained per channel.

What This Actually Changes for a Marketing Team

The practical shift is where the request originates, not what the agent does. If your team already lives in Claude for research, drafting, or thinking through a campaign, you can call an Orbitable agent from that conversation rather than switching tabs. The Dispatcher still does the orchestration work behind the scenes. MCP just adds another front door.

This matters most for teams whose actual workflow is scattered: some work happens in Slack, some in a CRM, some in whatever chat interface a team member has open. MCP doesn't collapse those tools into one place. It lets the agent fleet meet you in more of them.

MCP vs a Traditional API Integration

Dimension	Traditional point-to-point API	MCP
Build effort	One custom integration per pairing of app and AI tool	One server implements the standard once, any compliant client can use it
Who has to build it	Each vendor pair, separately, often duplicated across the industry	The tool vendor builds a server once; the AI client just needs to speak MCP
Where the call originates	Inside the vendor's own app or a dedicated integration panel	From any MCP client, such as a Claude conversation, outside the app entirely
Context handling	Usually re-authenticated and re-scoped per integration	Orbitable's world model travels with the agent regardless of entry point
Maintenance burden	Grows with every new AI client that wants to connect	Stays flat; new MCP clients can connect without new integration work

Where This Sits Alongside Orbitable's Other Integrations

Orbitable already connects to HubSpot, Attio, Slack, and Apollo. Those are data integrations: they pull information in and push actions out to the systems where your revenue data lives. MCP is a different layer. It doesn't connect Orbitable to a data source, it exposes Orbitable's own agents as tools that an external AI client can call. A HubSpot integration answers "where does the data come from." The MCP server answers "how do I trigger the work in the first place, and from where."

FAQ

Is MCP something Orbitable invented?

No. MCP is an open standard published by Anthropic in November 2024, not a proprietary Orbitable feature. Orbitable built a server that implements the standard, which is why the 50-agent fleet can be called from Claude or any other MCP-compliant client rather than only from inside the Orbitable app.

Do I need to write code to use Orbitable's MCP server?

No agent-building code is required to use it once it's connected; you drive agents through natural-language requests in the client, the same way you'd talk to Claude normally. Setting up the connection itself is a one-time configuration step in the MCP client you're using, not an ongoing development task.

Does calling an agent through Claude use credits differently than using the Orbitable app?

No, credits meter the work an agent does, not the entry point it was triggered from. A task run through the MCP server draws down credits the same way the identical task would if triggered inside the Orbitable app, because the meter tracks the agent's output, not which surface asked for it.

Is the MCP server the same thing as the HubSpot, Attio, Slack, or Apollo integrations?

No, those four integrations connect Orbitable to your existing data and communication tools, pulling records in and pushing actions out. The MCP server does the opposite job: it exposes Orbitable's own 50 agents as tools an external AI client, like Claude, can call directly.

Can any AI assistant use Orbitable's MCP server, or only Claude?

Any MCP-compliant client can connect, not only Claude, because MCP is an open standard rather than a Claude-specific feature. Claude is simply the client Orbitable built and tested this against first, since Anthropic authored the protocol.