HubSpot integrations with AgentKit

There’s a lot of buzz around OpenAI's AgentKit right now, and it’s easy to see why. Businesses are dreaming of hooking up powerful AI to their core systems like HubSpot, imagining smart agents handling everything from sales chats to customer support tickets.

But while the dream is big, the reality of creating HubSpot integrations with AgentKit is a bit more complicated than just flipping a switch. It’s easy to get swept up in the excitement, but taking a practical look under the hood reveals some important details you should know before you dive in.

This guide will give you a clear, no-fluff look at what AgentKit is, how it actually connects with HubSpot, what you can do with it, and the major limitations every business needs to consider before committing time and money.

What are HubSpot integrations with AgentKit?

Let's get one thing straight right away. When we talk about "integration" here, we're not talking about browsing an app store and clicking "install." Building an integration with AgentKit means you’re creating a custom AI agent from scratch that can talk to HubSpot's API.

Understanding HubSpot

You probably already know HubSpot as the all-in-one platform that combines marketing, sales, and customer service tools. A huge part of its appeal is its massive ecosystem of apps that let you easily add new features. That plug-and-play experience is what most HubSpot users are used to, which can make the hands-on nature of AgentKit a bit of a shock.

Understanding OpenAI's AgentKit

AgentKit isn't a finished product, it's a toolkit for developers who want to build AI agents. It’s made for people who write code and gives them three main building blocks:

1. Agent Builder: A visual space where a developer can draw out an agent's logic, connect it to different tools, and design workflows with multiple steps.

2. ChatKit: A collection of pre-made UI components to help you embed a nice-looking chat agent right into your website or app.

3. Connector Registry: A central spot for managing how your agent connects to other tools and data sources, including platforms like HubSpot.

How HubSpot integrations with AgentKit work: Features and the setup reality

The main way to get AgentKit and HubSpot talking is to build a custom conversational agent that can perform actions through the HubSpot API. The process usually looks something like this: a customer types a question into a chat window (built with ChatKit), their message gets sent to the agent (designed in Agent Builder), and the agent then uses the HubSpot API to either pull information (like a contact's details) or do something (like create a new support ticket).

Here’s the catch: this is not a simple setup. It takes real technical skill. A developer has to set up the API connections, often using something called a Model Context Protocol (MCP) server, and then write the actual code that tells the agent what to do with the info it gets from HubSpot.

graph TD  

    A[Customer asks question in Chat Window] -->|Built with ChatKit| B(Message sent to AI Agent);
    B -->|Designed in Agent Builder| C{Agent processes request};
    C -->|Uses custom code & MCP Server| D[Connects to HubSpot API];
    D --> E[Pulls contact info OR Creates support ticket];  

    E --> F[Sends response back to Customer];  

This is a completely different experience than what most teams are used to. For instance, platforms like eesel AI are designed to be user-friendly, offering one-click integrations with help desks and knowledge bases. You can be up and running in minutes, not months, without needing a developer on call.

Common use cases for HubSpot integrations with AgentKit

If you have the development team for it, you can build some genuinely useful tools. Here are a few things that are possible with a custom-built integration between HubSpot and AgentKit.

Building custom customer support chatbots

You could create an agent that answers customer questions by digging into your HubSpot knowledge base. If it gets stuck, it could automatically create a new ticket in HubSpot Service Hub and attach the full chat history for a human agent to pick up.

Creating intelligent sales development agents

Picture a chatbot sitting on your pricing page. It could ask a few qualifying questions to figure out a visitor's needs and budget. If they look like a strong lead, the agent could create a new contact and deal right inside your HubSpot CRM, maybe even assigning it to the right sales rep based on their territory.

Developing internal data assistants

You could have an internal agent on your company intranet that lets your team ask questions about HubSpot data in plain English. A manager could ask, "Show me all open support tickets for Company X," or "What were our sales last quarter?" and get an

answer instantly, without having to mess around with building a report.

While these custom agents are powerful, they require a lot of time and specialized knowledge. In contrast, tools like eesel AI offer AI agents for customer support and internal questions that are ready to go. They can be trained on your HubSpot data and other sources almost instantly, giving you value without the long wait.

Key limitations and challenges of HubSpot integrations with AgentKit

It's important to have a clear-eyed view of things. AgentKit is an interesting toolkit, but it comes with some big challenges that might make it the wrong choice for many businesses.

The challenge: It's a toolkit, not a ready-to-use platform

This is probably the most important thing to remember. AgentKit is made for developers. That means your non-technical folks, like support managers or sales ops leaders, can't build or even tweak these agents on their own. Every little change or new idea has to go through the engineering team, which creates bottlenecks and slows everything down.

This is a world away from a self-serve platform like eesel AI, which is built for the people who actually run your support team. You can set up and deploy AI, see how it would have handled past tickets, and gradually roll out automation with complete control, all without writing a single line of code.

Feature	HubSpot integrations with AgentKit	Self-Serve AI Platform (e.g., eesel AI)
Primary User	Developers, Engineers	Non-technical teams (Support, Sales)
Setup Time	Weeks to months	Minutes to hours
Maintenance	Requires ongoing developer support	Managed by business users
Flexibility	Highly customizable, but complex	Purpose-built, easy to configure
Time to Value	Slow, requires significant investment	Fast, immediate ROI

Vendor lock-in and unpredictable costs

When you build with AgentKit, you're tying your wagon to OpenAI's entire ecosystem. You're stuck with their models and their API. You can't just switch to a different model from a provider like Claude or Gemini if it happens to become cheaper or perform better for a certain task. That lack of flexibility can be a serious strategic risk down the road.

On top of that, the pricing is based on how many "tokens" you use, which can be hard to predict. A busy month for your support team could lead to a surprisingly large bill, making it tough to manage costs as you grow. By comparison, eesel AI has transparent and predictable pricing plans. They don’t charge you per resolution, so your costs won't balloon just because your automation is doing its job well.

Designed for chat, not autonomous automation

AgentKit is built from the ground up for chat-based conversations where a human is always involved. It’s great for building customer support chatbots that talk to people, but it’s not meant for automated workflows that run silently in the background. For example, it’s not the right tool for automatically sorting, tagging, and routing a new support ticket the second it comes in.

This is where dedicated AI support platforms have a clear advantage. For example, eesel AI's AI Triage product is built for exactly this purpose. It works behind the scenes to keep your support queues organized by routing, tagging, and handling tickets based on rules you set.

Data control and security risks

Since AgentKit is a cloud-only platform hosted entirely by OpenAI, it might not pass the strict data residency or security requirements that many larger companies have, especially those in regulated fields like finance or healthcare. You get very little control over the infrastructure where your customer data is being processed.

eesel AI was built with enterprise security in mind from day one. It offers EU data residency options and guarantees that your company’s data is never used to train general AI models. Your information remains your own, protected by enterprise-grade security and compliance.

Pricing considerations for HubSpot integrations with AgentKit

When you're trying to figure out the cost, you have to look at both sides of the equation.

• HubSpot: HubSpot's pricing is broken into different tiers for its Hubs, like Service Hub or Sales Hub. Plans usually go from Starter to Professional and Enterprise, and your level of API access can change depending on which plan you have. You can find the latest info on the official HubSpot pricing page.

• OpenAI AgentKit: AgentKit itself is free, but the cost comes from using the underlying OpenAI API, which is billed per token. Every time your agent processes a request or gives a response, it uses up tokens from models like GPT-4o. This means your costs are directly tied to usage, making them swing up or down from month to month. You can check the current rates on the official OpenAI API pricing page.

Are HubSpot integrations with AgentKit the right choice for your HubSpot stack?

So, what's the verdict? Should you use AgentKit to build on top of HubSpot? It all comes down to a simple trade-off. HubSpot integrations with AgentKit give you a ton of flexibility if you have a dedicated developer and AI team and want to build a completely custom chat experience from the ground up.

However, for most businesses, it’s a difficult and expensive route. The heavy reliance on developers, the risk of getting locked into one vendor, the unpredictable costs, and the narrow focus on chat instead of true automation make it a tough tool to get a quick return on.

If your goal is to get AI working for you quickly, automating support, helping your agents, and bringing all your knowledge from HubSpot and other tools together, a purpose-built platform is a much faster path to getting your money's worth. eesel AI offers a powerful, secure, self-serve solution that you can get running in minutes, not months.

This video from OpenAI provides a developer-focused demo on how to design, deploy, and optimize agentic workflows using AgentKit.