How to Connect HubSpot to AI Agents with MCP | Vantage Point

Introduction

Your CRM holds some of the most valuable data in your organization — customer histories, deal pipelines, support tickets, marketing engagement metrics. But accessing that data typically means logging into HubSpot, navigating dashboards, clicking through menus, and manually pulling reports.

What if your AI assistant could do all of that for you with a simple conversation?

That's exactly what the Model Context Protocol (MCP) makes possible. Developed as an open standard by Anthropic and now adopted across the AI ecosystem, MCP creates a secure bridge between AI agents and external systems like HubSpot CRM. Instead of building complex custom integrations, MCP provides a standardized way for AI tools to read, search, and manage your CRM data — all through natural language.

In this guide, we'll walk you through everything you need to know about connecting HubSpot to AI agents with MCP: what it is, how it works, step-by-step setup instructions, real-world use cases, and best practices for getting the most out of this powerful integration.

What Is MCP (Model Context Protocol)?

The "USB Port" for AI

Think of MCP as a universal adapter — just as USB standardized how devices connect to computers, MCP standardizes how AI models connect to business applications. Before MCP, connecting an AI agent to HubSpot required custom API development, authentication handling, and data formatting for each specific use case. MCP eliminates that complexity with an open, standardized protocol.

At its core, MCP defines three key components:

• MCP Servers — Applications (like HubSpot) that expose their data and capabilities through the protocol
• MCP Clients — AI tools (like Claude Desktop, Cursor, or ChatGPT) that connect to MCP servers to access data
• Tools — Specific actions the server makes available, such as searching contacts, updating deals, or creating tasks

When your AI assistant needs CRM data, it sends a structured request through MCP, the HubSpot MCP server processes it, and the response comes back — all in real time, all respecting your existing permissions and security settings.

Why MCP Matters for Your Business

Before MCP, organizations had three options for AI-CRM integration:

1. Manual copy-paste — Slow, error-prone, and impossible to scale
2. Custom API integrations — Expensive to build and maintain
3. Pre-built connectors — Limited in scope and flexibility

MCP offers a fourth path that combines the best of all worlds: standardized, secure, flexible, and increasingly supported across the entire AI ecosystem. As of 2026, HubSpot's remote MCP server is generally available, and major AI platforms including Claude, ChatGPT, Cursor, and Windsurf all support MCP connections.

How Does HubSpot's MCP Server Work?

HubSpot offers two distinct MCP servers, each designed for different use cases:

HubSpot MCP Server (Remote)

This is the primary server for business users and teams. Hosted at mcp.hubspot.com, it enables any MCP-compatible AI tool to securely access your HubSpot CRM data. Key capabilities include:

Read and Write Access:

• CRM records: contacts, companies, deals, tickets, products, line items, quotes, invoices, orders, carts, and subscriptions
• Activities: calls, emails, meetings, notes, and tasks
• Segments (lists)

Read-Only Access:

• Content and marketing: blog posts, landing pages, site pages, campaigns, and marketing events
• Organizational context: users, teams, owners, roles

Developer MCP Server (Local)

This CLI-based server is designed for developers building apps and CMS content on the HubSpot Developer Platform. It enables agentic development tools like Cursor and Windsurf to scaffold projects, deploy changes, and interact with HubSpot's development environment.

How a Typical MCP Interaction Works

Here's what happens when you ask your AI assistant, "Show me all open deals over $50,000 closing this quarter":

1. You prompt your AI assistant in natural language
2. The AI client translates your request into an MCP tool call (e.g., search_crm_objects)
3. The MCP server authenticates the request using your OAuth credentials
4. HubSpot's API processes the query, respecting your user permissions
5. Results return through MCP to your AI assistant
6. The AI formats the response in a clear, conversational summary

The entire round trip happens in seconds, and every action respects your existing HubSpot user permissions — users can only view and modify records they have access to.

Step-by-Step: Setting Up the HubSpot MCP Server

Prerequisites

Before you begin, ensure you have:

• An active HubSpot account (any tier — Free CRM works for testing)
• Access to the HubSpot Developer Platform (navigate to Development in your HubSpot account)
• An MCP-compatible AI client (Claude Desktop, Cursor, Windsurf, ChatGPT, or a custom application)

Step 1: Create an MCP Auth App in HubSpot

1. Log in to your HubSpot account
2. Navigate to Development in the main menu
3. In the left sidebar, click MCP Auth Apps
4. Click Create MCP auth app in the upper right
5. Fill in your app details:
   • App name: A descriptive name (e.g., "AI CRM Assistant")
   • Description: Brief description of how you'll use the connection
   • Redirect URL: The callback URL for your MCP client
   • Icon: Optional branding
6. Click Create

HubSpot will generate OAuth credentials (Client ID and Client Secret) — keep these accessible for the next steps.

Important: You don't manually define scopes for MCP auth apps. Available scopes are automatically determined by the tools available in the MCP server and the permissions each user grants during installation.

Step 2: Configure Your MCP Client

Configuration varies by client. Here are the most popular options:

Claude Desktop: Open Settings → Developer → Edit Config, and add the HubSpot MCP server configuration with your Client ID, Client Secret, and the server URL https://mcp.hubspot.com.

Cursor / Windsurf: Navigate to Settings → Tools & MCP → Add MCP Server. Enter the server URL and OAuth credentials. These IDE-based clients handle PKCE authentication automatically.

ChatGPT (via HubSpot Connector): HubSpot provides a native connector for ChatGPT that uses MCP under the hood. Install it from HubSpot's AI tools page.

Custom Applications: For custom integrations, your application must implement OAuth 2.1 with PKCE (Proof Key for Code Exchange).

Step 3: Authenticate and Connect

1. Initiate the OAuth flow from your MCP client
2. Select the HubSpot account you want to connect
3. Review and grant the requested permissions
4. Authorize the connection

Once authorized, your MCP client can begin making requests to the HubSpot MCP server.

Step 4: Test Your Connection

Start with a simple test prompt to verify everything works:

• "Get my user details and account information" — This calls the get_user_details tool
• "Search for contacts at [Company Name]" — This tests search_crm_objects
• "What deals are in my pipeline?" — This verifies CRM data access

Top Use Cases: What Can AI Agents Do with HubSpot MCP?

Sales Automation

• Pipeline management: "Show me all deals in the negotiation stage that haven't been updated in 14 days"
• Lead research: "Pull up everything we know about Acme Corp — contacts, deal history, and recent activities"
• Activity logging: "Log a meeting note for today's call with Sarah at TechStart — discussed timeline and budget"
• Forecasting: "Summarize all deals with close dates this quarter, grouped by stage and total value"

Marketing Intelligence

• Campaign analysis: "What were the top-performing campaigns last month by contact attribution?"
• Content insights: "Which blog posts generated the most form submissions in Q1?"
• Lead scoring: "Find contacts who visited our pricing page more than three times but haven't been assigned to a sales rep"

Customer Service

• Ticket triage: "Show me all open tickets older than 48 hours that don't have an assigned owner"
• Account context: "Before my call with DataFlow Inc., summarize their recent support tickets and deal status"
• Proactive outreach: "Which customers have submitted more than three tickets this month? Flag them for a relationship review"

Operations and Reporting

• Data quality: "Find contacts with missing email addresses or phone numbers in our active deals"
• Cross-object analysis: "Which companies have open deals but no recent activity logged in the past 30 days?"
• Bulk updates: "Create follow-up tasks for all deals closing this week that don't already have a scheduled meeting"

Security, Permissions, and Compliance

Built-In Permission Enforcement

Every MCP interaction respects your existing HubSpot user permissions. If a user doesn't have access to view financial data in HubSpot, their AI assistant can't access it through MCP either. This permission model means:

• Role-based access is automatically enforced
• Object-level permissions carry over to AI interactions
• Field-level security is respected

Sensitive Data Protection

HubSpot's MCP server does not allow access to custom Sensitive Data Properties, including Personal Health Information (PHI) and other forms of Highly Sensitive Data. If your account has Sensitive Data mode enabled, activity objects (calls, emails, meetings, notes, tasks) are blocked from MCP access entirely.

OAuth 2.1 and PKCE

The authentication layer uses OAuth 2.1 with PKCE, providing:

• No stored passwords — Access is token-based with automatic expiration
• Refresh token rotation — Single-use refresh tokens for enhanced security
• Granular revocation — Disconnect AI access at any time from HubSpot settings
• Audit trail — All actions are logged and attributable to specific users

Best Practices for Secure MCP Deployment

1. Start with read-only access — Grant write permissions only after thorough testing
2. Use sandbox environments — Test in HubSpot's sandbox before connecting to production data
3. Review permissions regularly — Audit which users and apps have MCP access quarterly
4. Monitor activity logs — Track AI-initiated changes alongside manual user activity
5. Implement least-privilege access — Only grant the scopes each use case actually requires

Best Practices for Getting the Most from HubSpot MCP

1. Craft Specific Prompts

Generic prompts produce generic results. Instead of "Show me my contacts," try:

• "Find all contacts tagged as 'Decision Maker' at companies with open deals over $25,000"
• "Search for contacts who were created this month and haven't received a follow-up email yet"

2. Combine MCP with Existing Workflows

MCP works best as an enhancement to your existing HubSpot processes, not a replacement. Use it to:

• Speed up data entry by logging activities conversationally
• Get quick pipeline summaries before meetings
• Identify at-risk deals that need attention

3. Test in Sandbox First

HubSpot offers sandbox environments where you can safely test MCP interactions without affecting production data.

4. Stay Current with Server Updates

HubSpot continuously adds new tools and capabilities to the MCP server. When new scopes become available, existing users need to re-install the app to grant them.

5. Plan for Token Management

OAuth access tokens expire after a set period. Ensure your implementation handles automatic token refresh, graceful re-authentication, and proper error handling for authentication failures.

How Vantage Point Helps You Implement HubSpot MCP

As a certified HubSpot partner, Vantage Point helps organizations implement and optimize AI-powered CRM workflows, including HubSpot MCP integrations. Our team specializes in:

• MCP Strategy & Architecture — Designing AI agent workflows that align with your business processes
• Custom Implementation — Building production-ready MCP integrations with proper security, error handling, and monitoring
• HubSpot Optimization — Ensuring your CRM data structure, permissions, and workflows are MCP-ready
• AI Agent Development — Creating custom AI agents that leverage MCP alongside Anthropic's Claude, Salesforce Agentforce, and other platforms
• Training & Enablement — Helping your teams adopt AI-assisted CRM workflows effectively

Frequently Asked Questions

What is the HubSpot MCP server?

The HubSpot MCP server is a Model Context Protocol endpoint hosted at mcp.hubspot.com that enables AI assistants like Claude, ChatGPT, and custom-built agents to securely read and write CRM data through natural language. It acts as a bridge between AI systems and HubSpot's APIs, translating conversational requests into API calls.

Do I need a paid HubSpot plan to use MCP?

No. The HubSpot MCP server works with any HubSpot tier, including the free CRM. However, the data and features accessible through MCP depend on your HubSpot subscription level and user permissions.

Which AI tools support HubSpot MCP?

As of 2026, major MCP-compatible clients include Claude Desktop, ChatGPT (via HubSpot's connector), Cursor, Windsurf, and any custom application that supports the MCP specification with OAuth 2.1 and PKCE authentication.

Is my CRM data safe when using MCP?

Yes. MCP enforces your existing HubSpot user permissions, uses OAuth 2.1 with PKCE for authentication, and does not expose sensitive data properties. All access is token-based with automatic expiration, and you can revoke AI access at any time.

Can AI agents modify my HubSpot data through MCP?

Yes. The MCP server supports write access for CRM records (contacts, companies, deals, tickets, line items, products) and activities (calls, emails, meetings, notes, tasks). Write operations respect user permissions, so only authorized users can make changes.

How is MCP different from HubSpot's standard API?

MCP adds a natural language layer on top of HubSpot's APIs. Instead of writing code to make API calls, your AI assistant handles the translation. MCP uses the CRM Search API under the hood but makes it accessible through conversation, dramatically lowering the technical barrier for CRM automation.

What happens if my access token expires?

OAuth access tokens expire after a set period. Your MCP client should automatically use the refresh token to obtain a new access token. If the refresh token has also expired, you'll need to re-run the full OAuth authorization flow.

Conclusion

The Model Context Protocol represents a fundamental shift in how businesses interact with their CRM. By connecting HubSpot to AI agents through MCP, you can transform your CRM from a tool you manually navigate into a platform your AI actively works with — querying data, logging activities, updating records, and surfacing insights through natural conversation.

The setup is straightforward, the security model is robust, and the use cases span every customer-facing function in your organization. Whether you're a sales leader looking to accelerate pipeline management, a marketer seeking deeper campaign insights, or an operations team focused on data quality, MCP-powered AI agents can make your HubSpot investment work harder.

Ready to connect your HubSpot CRM to AI agents? Contact Vantage Point to discuss your MCP implementation strategy and start building smarter, AI-powered CRM workflows today.

About Vantage Point

Vantage Point is a certified Salesforce and HubSpot partner specializing in CRM implementation, AI-powered automation, and system integration. We help businesses of all sizes unlock the full potential of their technology investments through expert consulting, custom development, and strategic guidance. Our team combines deep platform expertise with partnerships spanning Anthropic (Claude AI), Aircall, MuleSoft, and Workato to deliver comprehensive solutions that drive measurable results. Learn more at vantagepoint.io.