How to Connect ChatGPT to HubSpot CRM Using MCP

Updated for 2026: HubSpot now provides a generally available Remote MCP Server with read and write access to supported CRM data. HubSpot also provides an official HubSpot app for ChatGPT, so most users no longer need to manually build a custom MCP connection just to use HubSpot inside ChatGPT.

If you need more control, are developing your own AI integration, or specifically want ChatGPT Developer Mode to connect directly to HubSpot’s Remote MCP Server, you can still create a HubSpot MCP Auth App and connect it to https://mcp.hubspot.com.

This guide explains both approaches and, importantly, the difference between HubSpot’s Remote MCP Server and its separate Developer MCP Server.

Quick Answer: Can You Connect ChatGPT to HubSpot Using MCP?

Yes. HubSpot’s Remote MCP Server allows MCP-compatible AI clients to securely interact with supported HubSpot CRM data using OAuth 2.1 with PKCE. In 2026, the easiest option for most ChatGPT users is HubSpot’s official ChatGPT app. If you need a custom implementation, you can create a HubSpot MCP Auth App and connect ChatGPT Developer Mode directly to HubSpot’s remote MCP endpoint at https://mcp.hubspot.com.

TL;DR

HubSpot’s Remote MCP Server connects AI clients to real HubSpot CRM data.
For most users, the official HubSpot app in ChatGPT is now the simplest setup.
For custom integrations, HubSpot provides MCP Auth Apps and the remote endpoint https://mcp.hubspot.com.
Authentication uses OAuth and HubSpot’s MCP connection requires PKCE.
Supported operations include reading CRM records and creating/updating selected CRM objects and activities.
Access is limited by both the permissions granted to the connection and the connected user’s existing HubSpot permissions.

What Is MCP?

Model Context Protocol (MCP) is an open standard that allows AI applications such as ChatGPT to connect to external data sources and tools through a consistent interface.

Instead of teaching ChatGPT the details of every HubSpot API endpoint individually, an MCP server exposes defined tools that the AI client can discover and use.

For example, an MCP tool might allow ChatGPT to search HubSpot deals, retrieve a company record, create a task, or update a contact.

The official Model Context Protocol documentation describes MCP as a standardized way for AI applications to connect to external systems, data sources, tools, and workflows.

Why use MCP instead of building a HubSpot API integration from scratch?

MCP is useful when your main goal is conversational interaction. ChatGPT can discover available tools and decide which one to call based on your request.

A traditional API integration is still better when you need deterministic automation, webhooks, complex business logic, scheduled processing, bulk synchronization, or complete control over every API request.

HubSpot Remote MCP Server vs Developer MCP Server

This distinction is important because HubSpot has two different MCP servers.

Server	Purpose	Typical Use
HubSpot Remote MCP Server	Access HubSpot CRM data and supported actions	ChatGPT, AI agents, CRM querying and CRM updates
HubSpot Developer MCP Server	Work with HubSpot’s developer platform locally	Apps, UI extensions, CMS assets, CLI and development workflows

The Remote MCP Server is the one relevant when you want to connect ChatGPT to HubSpot CRM.

The Developer MCP Server runs locally and is installed through the HubSpot CLI. It is intended for development tools such as Codex CLI, Cursor, VS Code and other supported coding clients.

HubSpot explicitly documents these as separate products. See the Remote MCP Server documentation and Developer MCP Server documentation.

How ChatGPT + HubSpot MCP Works

The basic architecture looks like this:

User → ChatGPT → MCP Client → HubSpot Remote MCP Server → HubSpot CRM

Imagine asking:

“Show me all open HubSpot deals over $50,000 closing this month.”

ChatGPT interprets your request and selects an appropriate HubSpot MCP tool. The HubSpot MCP Server searches the CRM using the authenticated user’s access, returns the relevant data, and ChatGPT summarizes the results.

You do not need to know which HubSpot REST API endpoint or filter syntax is required.

What You Need Before Starting

The requirements depend on which connection method you use.

Requirement	Official HubSpot App	Custom MCP Connection
HubSpot account	Yes	Yes
Eligible HubSpot tier	Available across HubSpot tiers	Remote MCP Server is GA for HubSpot accounts
ChatGPT account	Yes; availability can depend on plan, region and workspace controls	Developer Mode is documented for Plus, Pro, Business, Enterprise and Education on web
ChatGPT Developer Mode	No	Yes
HubSpot MCP Auth App	No	Yes
OAuth / PKCE	Handled through the managed connection	Required

For HubSpot, the first installation of the official ChatGPT connector requires a Super Admin or a user with App Marketplace permissions. Individual users remain limited by their own HubSpot CRM permissions.

How to Connect ChatGPT to HubSpot Using MCP: Step by Step

Method 1: Use the Official HubSpot App in ChatGPT

This is the recommended method if your goal is simply to access and work with HubSpot from ChatGPT. You do not need to create your own MCP Auth App.

Step 1: Open ChatGPT Apps

In ChatGPT, open your profile menu and go to:

Settings → Apps

As of July 9, 2026, OpenAI uses the Plugins Directory as the main discovery experience for integrations, while connected apps are still managed under Settings → Apps.

Step 2: Find HubSpot

Open HubSpot from the available apps and click Connect.

You may also find HubSpot through the Plugins Directory depending on your ChatGPT interface and workspace configuration.

Step 3: Continue to HubSpot

Click Continue to HubSpot, sign in to HubSpot, and select the HubSpot account you want ChatGPT to access.

Step 4: Review HubSpot Permissions

HubSpot will show the permissions available to the connector. Review them carefully and enable only what your use case requires.

The first connection must be completed by a HubSpot Super Admin or a user with App Marketplace permissions.

Step 5: Complete the Connection

Click Connect App. HubSpot redirects you back to ChatGPT after authorization.

Step 6: Test the Integration

Start with simple read-only requests:

“Use HubSpot to find the company record for Acme Inc.”

“Show me my five most recently created HubSpot deals.”

“Summarize the open HubSpot deals in the Decision Maker Bought-In stage.”

“Summarize the recent calls, notes and meetings associated with this HubSpot contact.”

Step 7: Test a Write Action Carefully

The current HubSpot ChatGPT connector supports creating and updating several CRM records and logging activities.

For example:

“Create a follow-up task in HubSpot for this contact for next Monday.”

Review the proposed action before allowing ChatGPT to make the change.

HubSpot recommends configuring write tools so they require approval. OpenAI’s newer general app-permission settings also provide options that require confirmation before changes are made.

Advanced Method: Connect ChatGPT Directly to HubSpot’s Remote MCP Server

Use this method when you specifically want your own developer-mode MCP connection rather than HubSpot’s managed ChatGPT app.

Step 1: Enable Developer Mode in ChatGPT

OpenAI currently documents Developer Mode on ChatGPT web for Plus, Pro, Business, Enterprise and Education accounts.

For individual eligible accounts, navigate to:

Settings → Security and login → Developer mode

For managed Business, Enterprise or Education workspaces, administrator permissions and workspace controls may also apply.

See the OpenAI Developer Mode documentation.

Step 2: Start Creating a Developer-Mode MCP App

Open the ChatGPT Plugins area, select the plus button, and start creating a developer-mode app.

Your remote MCP endpoint is:

https://mcp.hubspot.com

HubSpot uses OAuth authentication, so the connection will also require HubSpot OAuth credentials and the exact redirect URL generated for your ChatGPT app.

Do not guess the redirect URL. OpenAI documents the current callback format as https://chatgpt.com/connector/oauth/{callback_id}, with the exact URL shown in the app management experience.

Step 3: Create an MCP Auth App in HubSpot

In HubSpot, go to:

Development → MCP Auth Apps → Create MCP auth app

Enter:

App name: for example, ChatGPT HubSpot MCP.
Description: a short internal description of the connection.
Redirect URL: the exact OAuth callback URL provided by ChatGPT.
Icon: optional.

After creation, HubSpot generates the OAuth client credentials for the MCP Auth App.

You can view the Client ID, Client Secret and redirect URL from the app details page.

Step 4: Configure OAuth in ChatGPT

Return to your ChatGPT developer-mode app and select OAuth authentication.

Use the credentials generated by the HubSpot MCP Auth App and ensure the configured redirect URL matches HubSpot exactly.

HubSpot requires OAuth authentication with PKCE for the Remote MCP Server. ChatGPT’s MCP OAuth implementation supports authorization-code authentication with PKCE.

Step 5: Scan the MCP Tools

Use ChatGPT’s Scan Tools option to inspect the tools exposed by the HubSpot MCP Server.

If OAuth authorization is requested, complete the HubSpot authorization flow and select the HubSpot account and permissions you want to grant.

After the scan completes, create or save the developer-mode app.

Step 6: Verify Permissions

HubSpot MCP Auth Apps work slightly differently from traditional HubSpot OAuth apps: you do not manually hard-code every scope when creating the MCP Auth App.

HubSpot determines the available permissions from:

The tools currently exposed by the HubSpot MCP Server.
The permissions selected by the user during installation.

The user’s existing HubSpot permissions apply on top of this. If a sales representative can only see certain deals in HubSpot, connecting ChatGPT does not give that representative access to every deal.

What Can ChatGPT Do With HubSpot After MCP Is Connected?

Use Case	Example Prompt	Access
Find contacts	“Find the HubSpot contact with email john@example.com.”	Read
Research companies	“Summarize the HubSpot company record for Acme Inc.”	Read
Analyze deals	“Summarize open deals above $25,000.”	Read
Review tickets	“Show the latest tickets associated with this customer.”	Read
Review engagement history	“Summarize the last three calls and notes for this account.”	Read
Create contacts	“Create this person as a HubSpot contact.”	Write
Update deals	“Update this HubSpot deal to the agreed stage.”	Write
Create tasks or notes	“Create a HubSpot follow-up task for this contact.”	Write
Analyze campaigns	“Summarize performance for this HubSpot campaign.”	Read

HubSpot’s Remote MCP documentation currently lists read access for contacts, companies, deals, tickets, users, carts, invoices, orders, line items, products, quotes, subscriptions, segments, activities, supported marketing/content data, and other documented sources.

Write access is currently more limited. HubSpot documents create/update capabilities for contacts, companies, deals, tickets, line items, products and supported activities such as calls, emails, meetings, notes and tasks.

Delete operations are not listed as supported CRM operations.

There is also an important difference between the generic Remote MCP Server and the managed ChatGPT connector. HubSpot’s Remote MCP documentation includes marketing events among readable marketing data, while the current HubSpot ChatGPT connector’s published object table does not explicitly list marketing events. If marketing-event access is critical, verify the available tool set before designing your workflow around it.

Practical ChatGPT + HubSpot Prompts

Sales

“Using HubSpot, show me open deals above $25,000 that have not had recent activity.”

RevOps

“Summarize our open HubSpot deals by stage and flag records that appear to be missing important deal information.”

Account Management

“Give me the latest HubSpot activity for Acme Inc. and summarize what has happened with the account.”

Support

“Summarize the most recent HubSpot tickets associated with this customer.”

Pipeline Review

“Using HubSpot, summarize deals in the Decision Maker Bought-In stage and group them by owner.”

Follow-Up

“Create a HubSpot task for me to follow up with this contact next Monday.”

Data Quality

“Review these HubSpot contacts and identify records with potentially incomplete or inconsistent information. Do not update anything.”

HubSpot MCP vs Traditional HubSpot API Integration

Area	HubSpot MCP	Traditional HubSpot API
Primary use	AI-driven conversational access	Application and system integrations
Natural-language interaction	Built for it	Requires your own AI/application layer
Development required	Low with managed connector; moderate for custom MCP	Usually higher
Control	Limited to exposed MCP tools	Much greater API-level control
Custom business logic	Limited	Excellent
Authentication	OAuth / MCP authorization	OAuth, app credentials or other supported HubSpot authentication
Best for	Research, CRM questions and human-in-the-loop actions	Production syncs, webhooks, bulk processes and deterministic automation

MCP should not be treated as a replacement for every HubSpot integration.

If you need to process thousands of records, react to webhooks, synchronize another platform continuously, implement strict business logic, or guarantee deterministic behavior, a traditional HubSpot API integration is usually a better architecture.

Security and Permissions

Connecting an AI assistant to CRM data should be handled like any other application integration.

Use OAuth and PKCE

HubSpot’s Remote MCP Server requires OAuth authentication with PKCE. PKCE helps protect the authorization-code flow against intercepted authorization codes.

HubSpot Permissions Still Apply

The MCP connection does not bypass HubSpot permissions. Users can only access or modify records that their HubSpot user account is permitted to access.

Use Least Privilege

Only enable the HubSpot permissions that users actually need. A user who only needs pipeline analysis may not need write permissions.

Require Approval for Changes

For CRM write operations, configure ChatGPT so changes require approval whenever possible. Review the exact record, property and value before approving an update.

Be Careful With Prompt Injection

OpenAI specifically warns that MCP integrations can be exposed to prompt-injection risks. Untrusted content encountered by the model could attempt to influence tool usage.

This is another reason to require confirmation for meaningful write actions and avoid combining highly sensitive access with unnecessary external sources.

Understand Sensitive Data Restrictions

HubSpot does not expose custom Sensitive Data Properties through the MCP connection. If Sensitive Data is enabled in the HubSpot account, activity objects such as calls, emails, meetings, notes and tasks, as well as conversation data, can also be blocked from MCP access.

Disconnect Unused Access

Remove integrations that are no longer required. In ChatGPT, connected apps can be managed under Settings → Apps.

Common Problems and Troubleshooting

Problem	Likely Cause	Fix
Developer Mode is missing	Unsupported account/surface or workspace restriction	Use ChatGPT web, verify plan eligibility, and check workspace administrator controls.
OAuth authorization fails	Incorrect redirect URL or OAuth configuration	Make sure the ChatGPT callback URL exactly matches the redirect URL configured in the HubSpot MCP Auth App.
PKCE authentication error	MCP client is not correctly using PKCE	Use a client with proper OAuth/PKCE support. HubSpot requires S256 PKCE for its MCP OAuth flow.
Connection stops working later	Access or refresh token problem	Reauthenticate if the refresh token is expired or invalidated.
ChatGPT cannot see expected records	HubSpot user permissions or connector permissions	Check both the user’s HubSpot CRM access and the permissions granted during connection.
New MCP capabilities are missing	Connection was authorized before new scopes/tools became available	Disconnect and reconnect so the updated permissions can be granted.
Write actions are unavailable	Read-only object, outdated connector or workspace action restriction	Verify the object supports writes, reconnect/upgrade the HubSpot app, and check ChatGPT workspace action controls.
Custom objects are unavailable	Not currently exposed by the managed HubSpot ChatGPT connector	Use the HubSpot API or another supported integration architecture when custom-object access is required.
Engagement activity is missing	Sensitive Data is enabled	HubSpot intentionally blocks MCP access to activity data in Sensitive Data-enabled accounts.
More than 10 records cannot be updated at once	Managed ChatGPT connector limit	The current HubSpot ChatGPT connector limits bulk create/update operations to 10 records per request.

Also note that HubSpot currently states that custom validation rules, including some pipeline-stage and association-label validations, are not applied when records are created or updated through the managed ChatGPT connector. Review important updates before approving them.

Who Should Use This Integration?

HubSpot consultants can use MCP to investigate CRM data, summarize account activity and answer client questions without manually building one-off API queries.

RevOps teams can use it for pipeline reviews, data-quality investigation and operational analysis.

Sales operations teams can inspect deals, companies and engagement history and perform controlled CRM updates.

CRM administrators can use it as a conversational layer over supported CRM records while keeping HubSpot’s existing user permissions in place.

Developers can create custom MCP-based experiences when the managed HubSpot ChatGPT app does not provide enough flexibility.

MCP is less suitable when you need scheduled automation, complex workflow logic, large-scale data synchronization, custom-object-heavy integrations, event-driven processing or guaranteed deterministic behavior. In those situations, consider HubSpot workflows, HubSpot APIs, a custom app, Breeze, or middleware such as Make, Zapier or n8n.

Final Thoughts

Connecting ChatGPT to HubSpot CRM using MCP is much more practical in 2026 than it was during the early MCP beta releases.

For most HubSpot users, the official HubSpot app in ChatGPT is now the best place to start. It removes most of the manual MCP configuration while providing access to supported CRM records and write actions.

The custom HubSpot Remote MCP Server route is still valuable when you need your own MCP client or developer-mode application. In that case, use HubSpot’s MCP Auth Apps, connect to https://mcp.hubspot.com, and follow the OAuth 2.1 and PKCE requirements.

The main consideration is not simply whether ChatGPT can access HubSpot. It is deciding which users should have access, which actions they should be allowed to perform, and which workflows are appropriate for an AI-driven interface rather than a traditional integration.

Frequently Asked Questions

1. What is the HubSpot MCP Server?

The HubSpot Remote MCP Server is a HubSpot-hosted service that allows MCP-compatible AI clients to interact with supported HubSpot CRM data. It can expose CRM records, activities and selected marketing/content information through standardized MCP tools. As of April 2026, the Remote MCP Server is generally available and supports both read and selected create/update operations.

2. Can ChatGPT connect directly to HubSpot using MCP?

Yes. You can use HubSpot’s official ChatGPT app for the simplest setup, or create a custom developer-mode connection to HubSpot’s Remote MCP Server at https://mcp.hubspot.com. The custom route requires an MCP Auth App, OAuth credentials, PKCE-compatible authentication and ChatGPT Developer Mode.

3. Is HubSpot MCP free?

HubSpot does not document the Remote MCP Server as a separately priced MCP add-on, and it is generally available to HubSpot accounts. Your normal HubSpot subscription, ChatGPT plan/features and applicable usage limits still apply. Certain capabilities can also depend on permissions, workspace configuration and the products available in your HubSpot account.

4. Which HubSpot plans support the MCP Server?

HubSpot announced the Remote MCP Server as generally available to all HubSpot accounts in April 2026. HubSpot also states that its official ChatGPT connector is available across HubSpot tiers. However, the actual data a user can access still depends on HubSpot user permissions and the features available in that account.

5. Which ChatGPT plans support MCP and Developer Mode?

OpenAI’s current Developer Mode documentation lists Plus, Pro, Business, Enterprise and Education accounts on ChatGPT web as eligible. Managed Business, Enterprise and Education workspaces can have additional administrator, RBAC and publishing controls. Availability of individual apps can also depend on plan, region, role and workspace configuration.

6. Can ChatGPT update HubSpot records using MCP?

Yes. HubSpot’s current Remote MCP Server supports create/update operations for selected CRM objects, including contacts, companies, deals, tickets, line items and products, plus supported activities such as calls, emails, meetings, notes and tasks. Always review write operations before approving them, especially when changing pipeline or customer data.

7. Can ChatGPT create contacts or deals in HubSpot through MCP?

Yes. Both contacts and deals are currently among the CRM objects that HubSpot allows supported MCP clients to create and update. The official HubSpot ChatGPT connector can also create contacts and deals. The connected user still needs the appropriate HubSpot permissions, and workspace controls can restrict write actions.

8. Is connecting ChatGPT to HubSpot through MCP secure?

The integration uses OAuth authorization, PKCE and HubSpot’s existing user permissions, which provide important security controls. However, AI tool use still requires good governance. Use least privilege, require approval for meaningful changes, review proposed write actions, understand Sensitive Data restrictions and consider prompt-injection risk when combining CRM access with untrusted information.

9. Does HubSpot MCP use OAuth?

Yes. HubSpot’s current Remote MCP Server requires OAuth authentication with PKCE. During authorization, users select a HubSpot account and approve available permissions. HubSpot then limits MCP access based on those granted permissions and the authenticated user’s existing CRM permissions.

10. What is the difference between HubSpot MCP and the HubSpot API?

HubSpot MCP provides AI clients with discoverable tools that are convenient for natural-language CRM interaction. The HubSpot API gives developers more direct and granular programmatic control. MCP is useful for human-in-the-loop AI workflows, while the traditional API remains better for webhooks, bulk integrations, custom logic, scheduled synchronization and production applications.

11. What is the difference between the HubSpot Remote MCP Server and Developer MCP Server?

The Remote MCP Server connects AI clients to actual HubSpot CRM information and supported CRM actions. The Developer MCP Server is a separate local tool used with HubSpot’s CLI and developer platform to help build apps, CMS assets, UI extensions and other developer resources. You need the Remote MCP Server for ChatGPT CRM access.

12. Do I need to build my own MCP server to connect ChatGPT to HubSpot?

No. HubSpot already hosts the Remote MCP Server at https://mcp.hubspot.com, and HubSpot also provides a managed ChatGPT connector. You only need to create your own middleware or MCP server when you require additional business logic, unsupported data sources, custom objects, special transformations or tools that HubSpot’s existing MCP implementation does not expose.