The HubSpot MCP server lets AI assistants like Claude securely read and write your HubSpot CRM data through a single, standardized connection. Instead of building a custom integration for every AI tool, your team points a Model Context Protocol (MCP) client at HubSpot, scopes what it can see, and starts asking questions in plain language.
For most businesses, the practical question isn't "what is MCP?" It's "should we connect an AI agent to our CRM, what can it actually do, and how do we keep it governed?" This guide answers that directly, using HubSpot's official documentation, and is written for any company on HubSpot — not a single industry.
What it is: The HubSpot MCP server is a secure gateway that connects any MCP-compatible AI tool or agent (Claude, Cursor, and similar clients) to your HubSpot CRM so it can retrieve, create, and update records using natural language.
Who it matters for: Operations, RevOps, sales, marketing, and technical teams at any business that wants AI to query and act on CRM data without a custom one-off integration for each tool.
What decision it supports: Whether to use HubSpot's native MCP server, build a custom integration, or wait — and how to scope access safely before agents can change live records.
Why Vantage Point is relevant: As a HubSpot Gold Partner and Anthropic Registered CPN member, Vantage Point helps mid-market teams design scoped, governed AI-to-CRM connections that produce real workflow value instead of risk.
The HubSpot MCP server is a secure connection point that lets AI tools interact with your HubSpot CRM through the Model Context Protocol. MCP is an emerging open standard that gives large language models (LLMs) a consistent way to request information from — and take action in — external systems like HubSpot, without each tool having to learn HubSpot's specific APIs.
HubSpot actually offers two distinct MCP servers:
This post focuses on the Remote server, because that's what lets your business put AI on top of live CRM data. For the source details, see HubSpot's MCP server documentation and the Public Beta changelog announcement.
AI assistants are only as useful as the context they can reach. A chatbot that can't see your pipeline, accounts, or tickets gives generic answers. The MCP server closes that gap by giving an AI client real, permissioned access to your CRM.
The strategic shift is standardization. Before MCP, connecting each new AI tool to HubSpot meant a separate custom integration. MCP acts as an abstraction layer over traditional APIs, so one connection pattern works across compatible clients. That lowers integration overhead and makes it realistic to let AI both read CRM data and act on it.
The flip side: when an agent can create and update records, governance stops being optional. The risk isn't reading data — it's an AI writing the wrong thing to a live system. That's why scoping and review matter, a point we return to below.
At a high level, the Remote MCP server is a bridge between an authorized AI client and your HubSpot account:
mcp.hubspot.com endpoint from any MCP-compatible client (such as Claude or Cursor). HubSpot supports OAuth 2.0 today and is aligning to OAuth 2.1 — including PKCE and refresh token rotation.| Access level | What the MCP server can reach |
|---|---|
| Read and write | CRM objects: contacts, companies, deals, tickets, carts, products, orders, line items, invoices, quotes, subscriptions, and segments (lists). Engagements: calls, emails, meetings, notes, and tasks. |
| Read-only | Organizational context: users, teams, reporting structures, owners, roles, seats. Marketing/content: campaigns and metrics, landing pages, website pages, blog posts. |
| Not accessible | Custom Sensitive Data Properties, including Personal Health Information and other Highly Sensitive Data. |
Available MCP tools let an agent retrieve, create, update, and list objects and their properties; create and read associations; add tasks and notes; and even open specific screens in the HubSpot UI. HubSpot notes that the set of tools and data will expand over time.
These work across industries — the value is the same whether you sell software, services, or products:
The MCP server is fast and standardized, but it isn't the only option. Use this decision table to choose.
| Factor | HubSpot MCP server (native) | Custom integration / middleware |
|---|---|---|
| Best for | Conversational AI access to CRM data and simple actions | Complex business logic, multi-system orchestration, scheduled syncs |
| Speed to value | Fast — scope an app and connect a client | Slower — design, build, and maintain |
| Standardization | High — one pattern across MCP clients | Custom per integration |
| Control granularity | Scope-based, admin-gated, evolving | Fully custom controls and audit logic |
| Maturity | Public Beta (Remote server) | Mature API and iPaaS tooling (e.g., Workato) |
| Maintenance | Managed by HubSpot | Owned by your team or partner |
Choose the native MCP server if you want AI agents to query and act on HubSpot data quickly, with centrally scoped access, and you're comfortable operating on a beta capability.
Choose custom integration or middleware if you need to coordinate HubSpot with ERP, billing, or data platforms, enforce complex rules, or require detailed audit trails today. Many teams do both — MCP for conversational access, an iPaaS layer for system-to-system automation. Our system integration and data migration team helps decide where each fits.
Governance matters most when an agent can act, not just read. A few principles keep MCP access safe:
If your team needs a formal access and review model, our compliance and security solutions help define controls that fit AI-to-CRM access.
Vantage Point is a mid-market CRM consultancy with senior-only, US-based consultants and an employee-owned model. As a HubSpot Gold Partner, an Anthropic Registered CPN member, and a Workato partner, we help teams turn AI-to-CRM connections into governed, useful workflows — not science projects.
We can run a HubSpot AI / MCP readiness assessment that maps your highest-value use cases, defines least-privilege scopes, sets a human-in-the-loop review model, and decides where the native MCP server fits versus custom integration. From there, we help implement and adopt it across sales, service, and marketing using our VALUE Methodology.
If you're evaluating AI on top of HubSpot, start with our HubSpot services and AI-driven personalization and analytics offerings. For related reading, see how HubSpot's Claude AI connector updates CRM records and our guide to using the HubSpot connector in ChatGPT.
It is a secure gateway that connects MCP-compatible AI tools and agents to your HubSpot CRM. Through it, authorized AI clients can retrieve, create, and update CRM records using natural language. HubSpot offers a Remote server for CRM data access and a local Developer server for building on the platform.
The Remote HubSpot MCP server is in Public Beta (announced May 6, 2025). The local Developer MCP server, used for app and CMS development through the CLI, has reached general availability. Because the Remote server is beta, test in a sandbox and expect features to evolve.
It has read and write access to CRM objects (contacts, companies, deals, tickets, products, orders, quotes, invoices, subscriptions, and lists) and engagements (calls, emails, meetings, notes, tasks). It has read-only access to organizational context and marketing content, and it cannot access custom Sensitive Data Properties such as PHI.
You control access through a user-level app and its OAuth scopes, granting only the read or write permissions a use case needs. An account admin must connect first before other users can, which centralizes control. For write actions, keep a human review step because LLMs can make mistakes.
Any MCP-compatible client can connect, including Claude and Cursor. The client authenticates with OAuth credentials to HubSpot's MCP endpoint. HubSpot supports OAuth 2.0 today and is moving toward OAuth 2.1 with PKCE and refresh token rotation.
Use a custom integration or middleware when you need complex business logic, orchestration across systems like ERP or billing, scheduled syncs, or detailed audit trails today. The native MCP server is best for fast, standardized conversational access and simple actions. Many teams use both.
It can be, with disciplined governance. Use least-privilege scopes, gate onboarding through an admin, keep humans in the loop for writes, and test in a sandbox first. The server also excludes custom Sensitive Data Properties by design. Vantage Point can help define the access and review model.
Confirm you're on the new Developer Platform, pick one low-risk use case, create a user-level app with minimal scopes, have an admin connect, and test with a compatible MCP client in a sandbox. Add write access only after validating read behavior. A readiness assessment from Vantage Point can accelerate this safely.