Sales ops · Production

HubSpot builds a remote MCP server to expose CRM data to AI agents

The problem

HubSpot needed to connect their CRM to AI agents, with 75% of customers already using ChatGPT, but the MCP standard lacked built-in auth support and no agent protocol had yet emerged in the AI era, making it unclear how to expose CRM data securely at scale.

Workflow diagram · grounded in source
1
OpenAI MCP adoption triggers build
trigger
“there was a date that was a turning point. And it was the turning point for when OpenAI adopted MCP. And then you can see in the span of 48 hours, you saw all the big players Microsoft, Google, OpenAI, all adopt it”
2
Build vs. buy evaluation
human_review
“we looked at a whole bunch of different evaluations, build and buy, and we decided that the things that we needed, it would have been faster and easier to build. Because we needed it to work with our existing enterprise RPC system”
3
Extend Java MCP SDK
integration
“we could extend the Java MCP SDK to do things like HTTP streaming protocols”
4
Implement OAuth for CRM access control
validation
“what would be the fastest way for us to ensure that you can only see HubSpot CRM data that you're allowed to? We decided to use the OAuth protocol for that”
5
Build MCP gateway with auto-discovery
integration
“we built kind of an MCP gateway. It auto-discovers tools across those services. And then we needed to add the OAuth scope, and then the user permissions that were mapped to our HubSpot seat and tiers”
6
Claude Code generates tool annotations
ai_action
“we created an annotation system for MCP tools. So it was easier to describe these RPC Java methods. And that, we used Claude Code to generate them. That saved a lot of time”
7
CRM search API exposed via MCP
output
“One exposes our core raw APIs, our CRM search API with contacts, companies, and deals. It's built on top of our existing REST APIs”
Reported outcome

HubSpot became the first CRM with a remote MCP server and the first CRM connector to OpenAI, delivered in under four weeks, with internal AI tool adoption reaching 70-80% and customers actively using the connector.

Reported metrics
customer ChatGPT usage75%
Connector build timeless than four weeks
internal AI tool adoptionupwards in the 70-80%
third-party HubSpot MCP servers created by communityover 100
Reported stack
Java MCP SDKClaude CodeDropwizardOAuthBreeze Assistant
Source
https://stackoverflow.blog/2025/09/16/what-an-mcp-implementation-looks-like-at-a-crm-company/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

HubSpot became the first CRM with a remote MCP server and the first CRM connector to OpenAI, delivered in under four weeks, with internal AI tool adoption reaching 70-80% and customers actively using the connector.

What tools did this team use?

Java MCP SDK, Claude Code, Dropwizard, OAuth, Breeze Assistant.

What results were reported?

customer ChatGPT usage: 75%; Connector build time: less than four weeks; internal AI tool adoption: upwards in the 70-80%; third-party HubSpot MCP servers created by community: over 100 (source-reported, not independently verified).

How is this sales ops AI workflow structured?

OpenAI MCP adoption triggers build → Build vs. buy evaluation → Extend Java MCP SDK → Implement OAuth for CRM access control → Build MCP gateway with auto-discovery → Claude Code generates tool annotations → CRM search API exposed via MCP.

Related sales ops cases