How To Connect HubSpot (and your GTM strategy) to the World With MCP
Unlock powerful GTM workflows and campaigns with AI
👋 Hey, I'm Stuart, and welcome to ConnectedGTM. This edition has been made available for 💸 FREE 💸 by Flowla | Flowla 2.0 turns static sales rooms into dynamic workflows that improve prospect experience and remove sales rep busy work.
Back at Inbound 2022, HubSpot CEO Yamini Rangan introduced the connected customer platform, a vision that you could bring your entire customer journey into HubSpot. Fast forward to 2025 and HubSpot has made huge strides in that direction.
But there's always been a fundamental challenge when you’re trying to run a connected go-to-market motion:
HubSpot simply doesn't (and likely can’t) have ALL the context GTM teams need to make the most strategic decisions.
While the app marketplace and custom integrations solve part of the problem, they've always been limited by the rigid and structured nature of APIs and HubSpot’s data model, requiring you to know specific schemas and configure your CRM precisely.
This week, HubSpot released a new beta that changes this integration paradigm.
MCP (Model Context Protocol) - a bridge that allows LLMs like Claude to directly access and modify your HubSpot data through natural language conversation.
HubSpot has moved remarkably quickly to release their MCP Server, and I'm genuinely excited to see what it enables for go-to-market teams, so that’s what we’re digging into today.
This edition of ConnectedGTM was kindly supported by:
Your deal room shouldn't just look good, it should do the work. Flowla 2.0 turns static sales rooms into dynamic workflows that improve prospect experience and remove sales rep busy work.
Book a demo with the Flowla team to see the workflows you could build for your team.
Understanding MCP: The universal language to connect your tools with AI
Most companies are building GTM stacks that are disconnected by default. Even if you’re using best in class integrations, you can’t engage with ALL your integrated data through one pathway.
If you’re trying to use the power of AI to reason and analyze your data across sources you’re likely battling:
❌ Constant context switching as you jump between data and AI tools
❌ Outdated (or unverifiable) information as AI can't access your real-time CRM data
❌ Manual data entry when adding AI-generated insights back into HubSpot
What if your could talk directly to your HubSpot data and combine it with research from the public web (via AI Agents) and any other data source you like using an LLM, without you having to copy/paste or build complex integrations?
MCP is fundamentally different way to connect and engage with data, it isn’t limited by the constraints of traditional integrations:
Rigid API requirements and specific data formats
Complex authentication and permission management
Technical knowledge needed to build and maintain connections
Point-to-point integrations that don't scale easily
So what exactly is MCP?
MCP is an open standard that enables AI systems to communicate with software applications through a consistent language. In simple terms:
Think of MCP as a special mail service for robots 🤖
The AI is like a kid who writes a letter, and MCP makes sure every different system (like a library, a game, or a school database) can read that letter and answer back.
This way, the LLM can get answers or do things in lots of places using one simple mailbox system instead of many complicated ones.
By breaking down data silos, MCP helps advanced AI models give better, more relevant answers because they can fetch up-to-date facts and context.
How to connect to the HubSpot MCP server (plus a practical workflow example)
Let's walk through both setting up the MCP connection with an example of how to use this new connection to analyze our closed-won deals and generate a CRM verified segment and campaign idea.
Step 1: Create a Private App in HubSpot
Log in to your HubSpot account, navigate to Settings → Integrations → Private Apps and Click "Create private app"
Fill in the basic information:
→ Name: "MCP Access"
→ Description: "Connects HubSpot to Claude via Model Context Protocol"Configure auth by selecting the necessary scopes:
Contacts (Read, Write)
Companies (Read, Write)
Deals (Read, Write)
Create the app and copy your access token
💡 Security Tip: For initial testing, grant limited read-only permissions first. You can expand to write permissions once you're comfortable with how the system works.
Step 2: Connect Claude Desktop to the HubSpot MCP Server
Install and open the Claude Desktop app and click on settings and in the app toolbar (at the top of the screen not inside the app) click “Settings”
Select "Developer" in the left sidebar and then click “Edit Config”
This will open a JSON file on your computer “claude_desktop_config.json”. Paste the following code including your HubSpot access token from step 1 into that file.
{
"mcpServers": {
"hubspot": {
"command": "npx",
"args": ["-y", "@hubspot/mcp-server"],
"env": {
"PRIVATE_APP_ACCESS_TOKEN": "< YOUR HUBSPOT ACCESS TOKEN >"
}
}
}
}
Save the file and you should now see HubSpot and it’s 21 available tools inside of the “Search and Tools” menu for a new Claude chat.
If you’ve made it this far. You’re ready to “talk” to HubSpot.
Example HubSpot Workflow: Deal Pipeline Analysis and Action
Here's a complete workflow demonstrating how MCP lets you combine HubSpot data with external research to build a hyper-targeted outreach strategy:
Discover Common Traits in Your Best Customers
We all want better ways to refine our segmentation and targeting to be able to send the most relevant message so let’s start by asking Claude to analyze recent closed-won deals for patterns:
What are the common traits of my closed deals in HubSpot, research the companies in the last 5 closed won deals to find signals that could be used to create a tightly defined segment with a common acute pain point.
Claude will then:
Query HubSpot for your recent closed-won deals
Retrieve all associated company data
Analyze patterns across company size, industry, technologies used, etc.
Look for commonalities in the sales notes and interactions
The power of MCP here is that Claude can run multiple HubSpot queries to gather all this data without you having to specify exactly how to get each piece of information, it knows how to navigate HubSpot's data structure.
Once Claude analyzes the HubSpot data, it will present you with:
Common characteristics of your winning deals
Patterns in the problems these customers were trying to solve
The specific pain point that your solution addressed
For example, Claude might discover: "Your last five closed deals were all mid-sized manufacturing companies (100-500 employees) that mentioned 'inventory forecasting challenges' during the sales process.
Now comes the truly powerful part, instead of being limited to just HubSpot data, you can ask Claude to combine what we know from our CRM with other sources of data and build a more in depth understanding of those new customers.
Try prompting Claude:
What would be publicly available signals that a business might be experiencing this kind of acute pain point? Give me 10 to 20 signals and potential sources of the data that I might use to identify them.
Now you can use these recommendations (either directly in Claude with web search turned on), by calling a more specialized MCP connected data source, or by referencing additional data enriched in HubSpot to identify a list of additional companies that match the same profile as your recently closed won deals.
Now what I want you to do is to search HubSpot for a list of 20 additional companies that might meet these same criteria and be currently experiencing the same acute pain as the closed won deals you analyzed.
The result is a highly personalized, pain-focused segment based on a combination of your internal HubSpot data and external research all orchestrated through a simple conversation with Claude.
p.s. If you’re interested in learning more about how you can use AI to identify new segments and send more relevant campaigns in HubSpot I’d love to chat.