TL;DR Article Summary
HubSpot MCP lets MCP-compatible AI tools connect to HubSpot through a standard protocol instead of a one-off integration. HubSpot currently offers two MCP servers: a remote HubSpot MCP server for CRM data access and a local Developer MCP server for building apps and CMS assets. In order to use the remote server, create an MCP auth app in HubSpot, connect your MCP client to https://mcp.hubspot.com, authenticate with OAuth using PKCE, and start prompting against HubSpot data. Workato extends this pattern by exposing governed, multi-system business actions as MCP tools across the rest of the enterprise stack, including marketing, finance, support, data warehouses, and approval systems.
The popularity of AI has also brought about a change in the kind of use cases that folks work on day in, day out. Sales and marketing teams are also witnessing these changes.
A RevOps manager might want to know which high-value deals have gone quiet. A campaign manager might want to ask, in plain English, which HubSpot campaigns influenced the pipeline last quarter. None of these are distinct, futuristic use cases anymore.
One way to support these workflows is MCP.
HubSpot MCP gives MCP-compatible AI tools a standardized way to connect to HubSpot CRM data, so agents can query records, summarize activity, inspect deals, and, in supported cases, create or update CRM objects through authenticated tool calls.
You’d notice that a typical workflow will extend beyond HubSpot and venture into Slack, Google Ads, Snowflake, Zendesk, etc.
This is where Workato can help turn that context into governed, multi-system actions. For example, an agent might use HubSpot MCP to identify stalled deals, then use Workato to notify the right team in Slack, enrich the account from Snowflake, or create a follow-up workflow.
This guide explains how HubSpot MCP works, how to set it up, and how to extend it into broader enterprise workflows with Workato.
What Is HubSpot MCP?
Model Context Protocol (MCP) is an open standard that lets AI assistants and agents connect to external tools, systems, and data sources in a consistent way.
Instead of hard-coding every CRM action into an agent, MCP gives the agent a structured way to discover available tools and request data or actions.
HubSpot MCP refers to HubSpot’s MCP servers, which let compatible AI tools work with HubSpot either as a CRM data source or as a developer platform.
This means an AI assistant can help summarize deals, inspect CRM records, create or update supported records and activities, or support HubSpot app development, depending on which server you use.
In fact, HubSpot offers two different MCP experiences, which we’ll cover in the next section.
Remote HubSpot MCP Server vs. Developer MCP Server
The remote HubSpot MCP server is the one most RevOps teams, AI engineers, and CRM admins are looking for when they want to connect AI agents to HubSpot CRM data.
It is hosted by HubSpot at https://mcp.hubspot.com and supports secure read and write access to supported CRM data through OAuth-based authentication with PKCE.
The Developer MCP server is a local, CLI-based server for developers building HubSpot apps and CMS content. It runs through the HubSpot CLI and helps agentic coding tools interact with HubSpot’s developer platform.
You should use the remote server when you want an AI tool to address things like, “What changed on this deal?” or “Create a follow-up task for this contact.”
And you should use the Developer MCP server when you want help scaffolding a HubSpot UI extension, working with CMS assets, or navigating HubSpot developer tooling.
What Does the Remote HubSpot MCP Server Do?
The remote HubSpot MCP server acts as a bridge between an MCP-compatible AI client and HubSpot CRM data.
According to HubSpot’s documentation, it can expose supported CRM records, activities, marketing and content data, campaign data, and organizational context, while still respecting existing HubSpot user permissions.
Available capabilities include reading CRM records such as contacts, companies, deals, tickets, users, carts, invoices, orders, line items, products, quotes, subscriptions, and lists. It also supports activity data like calls, emails, meetings, notes, and tasks.
For write access, HubSpot documents support for contacts, companies, deals, tickets, line items, products, and activities such as calls, emails, meetings, notes, and tasks.
But there are some limits. HubSpot dictates that actions respect the authenticated user’s existing permissions.
It also notes that custom sensitive data properties are not available through the MCP server, and if “Sensitive Data” is turned on for an account, activity objects may be blocked through MCP even though standard CRM APIs may still support them.
What Does the Developer MCP Server Do?
The Developer MCP server is for local development workflows. It lets supported coding tools interact with the HubSpot Developer Platform through the HubSpot CLI.
You might use it to ask how a HubSpot CLI command works, scaffold a UI extension project, inspect developer documentation, or help manage local app and CMS development tasks.
During setup, you choose which supported clients to configure, such as Cursor, Claude Code, Codex CLI, Gemini CLI, VS Code, or Windsurf. HubSpot also supports a standalone mode that uses npx @hubspot/cli, which can be useful when global npm installs are restricted.
How to Set Up HubSpot MCP for CRM Access
For CRM data access, you should use the remote HubSpot MCP server. We’ll explore the process of setting up the MCP server in this section.
Create an MCP Auth App in HubSpot
In your HubSpot account, go to Development, then MCP Auth Apps. Create a new MCP auth app and enter the app name, description, redirect URL, and optional icon.
HubSpot then generates OAuth credentials for the app. If you are testing with MCP Inspector, HubSpot’s docs note that you should include http://localhost:6274/oauth/callback/debug as a redirect URL.
Configure Your MCP Client
Your MCP client needs to connect to https://mcp.hubspot.com, and you will need the client ID, client secret, and redirect URL from the MCP auth app details page. The redirect URL in the client must match the one configured in HubSpot.
Authenticate With OAuth Using PKCE
PKCE, or Proof Key for Code Exchange, is an OAuth security extension that helps protect authorization flows, especially for public clients.
HubSpot requires PKCE for MCP authentication, so your MCP client must either handle it automatically or implement the verifier and challenge flow correctly.
During authorization, the user selects the HubSpot account, grants permissions, and authorizes the app. The resulting access token lets the MCP client make HubSpot MCP requests on behalf of that user.
Test the Connection
For quick validation, HubSpot recommends MCP Inspector because it handles OAuth with PKCE automatically. Once connected, you can list tools and run a basic tool such as get_user_details to confirm the authenticated user, account, and available object access.
Start Prompting
Once your MCP client is connected, you can use natural language prompts to query and act on HubSpot data.
Example HubSpot MCP Prompts
It might be useful to check some sample MCP prompts to use with the HubSpot MCP server. You should start with low-risk, read-only prompts before allowing write actions.
Try prompts like:
- “Summarize all open deals over $25,000 that are currently in the proposal stage.”
- “Find contacts at Acme Corp and summarize recent activity.”
- “Which deals have no activity in the last 14 days?”
- “Show me the owners for deals expected to close this month.”
- “Find the property definitions for lifecycle stage and lead status.”
The best prompts are specific about object type, filters, date range, and desired output. “Summarize pipeline risk” is vague. “List enterprise deals over $50,000 closing this quarter with no logged meeting in the last 30 days” is much more useful.
Best Practices for Using HubSpot MCP
Every tool has some recommended usage guidelines, and that holds true for HubSpot MCP as well. Here are some of those guidelines:
1. Start With Read-Only Use Cases
It is important to let teams build trust by using AI to summarize deals, inspect records, and answer operational questions before enabling record creation or updates.
2. Use Least-Privilege Access
Since HubSpot MCP respects user permissions, you should make sure that HubSpot roles and permissions are clean before connecting agents.
3. Be Explicit in Prompts
Ask for the exact object, filter, field, or output format you want.
4. Monitor Write Actions Carefully
Creating tasks or updating records can save time, but it can also create bad CRM hygiene if prompts are vague.
5. Plan for Token Refresh
OAuth access tokens expire, so custom clients should use refresh tokens correctly and rerun authorization when needed.
6. Avoid Sensitive Data Workflows Unless You Have Verified Support
HubSpot places restrictions around sensitive data properties and activity access in sensitive-data-enabled accounts.
How Workato Extends MCP-Style Workflows Across the Enterprise
HubSpot MCP is powerful when the job begins and ends inside HubSpot. But many enterprise workflows do not. For example, a campaign performance question may require HubSpot, Google Ads, Snowflake, and finance data.
A sales follow-up workflow may need HubSpot, Slack, Gong, DocuSign, and an approval process.
This is where Workato extends the MCP pattern beyond a single application. Workato’s Enterprise MCP approach connects AI agents to governed workflows across hundreds of apps and systems, not just one CRM.
Using this, teams can expose reusable business actions as MCP tools, add approval steps, enforce rate limits, apply user-based access, and maintain audit trails.
For example, an AI agent could use HubSpot context to identify stalled deals, enrich the account with product usage data from a warehouse, notify the account team in Slack, and create a follow-up task only after checking business rules.
HubSpot provides CRM context, while Workato can expose governed, reusable actions that connect that context to broader enterprise workflows.
If your team is experimenting with HubSpot MCP, the next step is to explore how Workato can connect HubSpot data and AI agents to the rest of your business systems. That way, agentic workflows can move beyond CRM lookup and into secure, multi-system execution.
FAQs
Does HubSpot Have an MCP Server?
Yes. HubSpot offers two MCP servers: a remote HubSpot MCP server for CRM data access and a local Developer MCP server for app and CMS development.
The remote server connects MCP-compatible AI tools to HubSpot CRM data, while the Developer MCP server helps developers work with HubSpot’s developer platform through the CLI.
How Do I Set Up HubSpot MCP?
To set up the remote HubSpot MCP server, create an MCP auth app in HubSpot under Development > MCP Auth Apps.
Then configure your MCP client to connect to HubSpot MCP using the generated OAuth credentials. Your client must support OAuth with PKCE, and the redirect URL must match your HubSpot app configuration.
Is MCP an API Gateway?
MCP is not exactly an API gateway. It is a protocol that lets AI tools discover and use external tools in a standardized way. In HubSpot’s case, the remote MCP server acts as a bridge between AI clients and HubSpot APIs, but MCP itself is broader than a traditional API gateway.
What Is the Difference Between HubSpot MCP and the HubSpot API?
HubSpot APIs are designed for developers building direct integrations with explicit endpoints and payloads. HubSpot MCP gives AI tools a standardized interface for using HubSpot capabilities through natural language and tool calls.
For deterministic backend integrations, APIs may still be better; for agent-assisted workflows, MCP can be faster to adopt.
Can HubSpot MCP Update CRM Records?
Yes, the remote HubSpot MCP server supports write access for certain CRM records and activities. HubSpot documents write support for contacts, companies, deals, tickets, line items, products, calls, emails, meetings, notes, and tasks.
Access still depends on the authenticated user’s HubSpot permissions and the tools available through the MCP server.
This post was written by Deboshree Banerjee. Deboshree is a Senior Backend Engineer at Stripe and has previously held engineering roles at Goldman Sachs, Walmart, and tech lead roles at venture-backed startups. Throughout her career she has worked on large-scale distributed systems where data durability, redundancy, security, and operational reliability are critical concerns. Her experience evaluating enterprise software through the lens of scalability, resilience, and data protection informs her analysis of SaaS backup and recovery platforms.