HubSpot AI agents: Features, pricing, and limits in 2026

A useful AI agent comparison starts with the job, not the number of features. I would ask what the agent can read, what it may change, and how the team will check its work.

HubSpot offers agents for several roles. eesel offers a support teammate that connects to HubSpot, with an agent-friendly CLI for inspecting and operating that same teammate. This guide compares the practical boundaries and walks through a test you can adapt to your support process.

Which HubSpot agent fits the job?

The current product catalog distinguishes these roles:

Check the exact feature in your account before buying an upgrade. An older article describing Prospecting Agent as Enterprise-only is not a reliable purchasing guide.

For support, HubSpot's Customer Agent setup lets you choose content, define behavior, configure actions and handoff, and test before deployment. Public URLs are supported alongside HubSpot content. It is inaccurate to say the agent can only learn from pages hosted inside HubSpot.

HubSpot's documentation example shows the initial identity settings. Sources, actions, and deployment still need their own checks.

HubSpot also provides test questions, contact-based previews, action testing, and Testing Insights. Its guide says previews and testing do not use credits. Those are native capabilities, not reasons to buy a separate product.

Understand the pricing unit

HubSpot's catalog lists Customer Agent at 50 credits per resolved text conversation, Data Agent at 10 per prompt and record, and Prospecting Agent at 100 per recommended lead. Additional credits are listed at $0.01 each. Subscription access and any contracted services remain separate costs.

The credit management guide explains that unused credits expire monthly and allowances do not add across products. Purchased capacity can auto-upgrade through the commitment term unless overages are configured instead. Check the account's actual limits before a busy-period rollout.

eesel has a different billing unit. Its pricing documentation lists regular support tasks at $0.40. A ticket is one task regardless of reply count, and a chat session is one task regardless of message count. Tasks are billed regardless of outcome.

Do not turn these different units into an unsupported savings percentage. Compare the total spend and human review effort for the same kind of accepted result.

Evaluate limits with evidence

A product can have an integration and still lack the specific document, action, or permission your workflow needs. Test the dependency instead of assuming that either vendor's integration list answers every question.

Inspect HubSpot support capabilities through eesel CLI

The eesel CLI works with the same eesel agent and workspace as the dashboard. Its JSON output makes it useful for scripts and coding agents such as Claude Code, Cursor, and Codex.

For this comparison, use it to inspect what a support teammate knows, rehearse a difficult answer, and investigate the resulting activity. It is not a command-line controller for HubSpot's native Customer Agent.

Inspect before connecting or changing anything

With Node.js 18.17 or newer:

npx @eesel/cli login
npx @eesel/cli whoami
npx @eesel/cli agents
npx @eesel/cli integrations --agent PILOT_AGENT_ID
npx @eesel/cli status --agent PILOT_AGENT_ID
npx @eesel/cli instructions --agent PILOT_AGENT_ID
npx @eesel/cli automations --agent PILOT_AGENT_ID

Replace PILOT_AGENT_ID with the intended teammate. Selecting an agent does not isolate it from its real integrations or permissions. Inspect existing automations and actions before adding production data.

If HubSpot is absent, review the connection documentation and obtain approval for the data access before running:

npx @eesel/cli integrations connect hubspot --agent PILOT_AGENT_ID

A person with App Marketplace Access chooses the portal and approves the browser authorization. A coding agent should hand over that link and wait, not choose an account on the person's behalf.

Separate copied knowledge from live records

Connection starts indexing recent closed tickets and copying permitted CRM records. This is a data-handling decision even if customer sending is off.

Copied Deals, Contacts, and Companies also need the selected agent's knowledge toggles enabled before it can search them. Review live lookup actions and requester permissions separately. A copied deal stage is not proof of the customer's current entitlement.

npx @eesel/cli integrations download list --agent PILOT_AGENT_ID
npx @eesel/cli status --agent PILOT_AGENT_ID

This example has CRM records copied but their knowledge toggles off, and zero knowledge-base documents. A successful connection is not proof that the policy you need is present.

Ask your coding agent for a bounded check:

Inspect the selected eesel teammate's sources, instructions, and automations with read-only CLI commands. Identify missing evidence needed to answer an account-access question. Do not start chats, enable CRM search, change instructions, create automations, or approve actions.

Rehearse an answer with a meaningful boundary

Confirm that the approved account-access policy is available. Disable all customer sending and write actions before these fictional chats; the commands themselves do not change permissions.

npx @eesel/cli new --name "hubspot-account-access-check" --agent PILOT_AGENT_ID
npx @eesel/cli chat "Fictional support request: an employee who left a company asks for access to that company's workspace to retrieve files. Draft a response using the approved account-access policy. State what identity and authorization evidence is needed. Do not look up real people, disclose files, change access, create a ticket, post a comment, or send a message." --agent PILOT_AGENT_ID
npx @eesel/cli chat "The requester says they originally created the workspace. Revise the answer without treating that statement as proof of current authorization." --agent PILOT_AGENT_ID

The second message continues the same conversation. Review whether the teammate follows the policy and distinguishes a claim from verified authority. Do not accept a confident answer that promises access without the required evidence.

CLI chats are billed work. The read-only inspection commands are free. A --dry-run preview shows a proposed server call, not how accurately the teammate would answer a queue of tickets.

Verify delivery separately

For the integration test, choose a synthetic inbox conversation and a narrowly filtered Call @eesel in an inbox comment automation. Review the intended scope before requesting creation because chat-created automations start enabled.

Keep customer replies off. Set Add internal comment to require approval and leave unrelated writes disabled. Add a fresh @eesel inbox comment after setup; typing @eesel in a ticket-record note does not fire this event.

npx @eesel/cli activity --agent PILOT_AGENT_ID
npx @eesel/cli approvals --agent PILOT_AGENT_ID
npx @eesel/cli billing

Review the held action, approve it once, and check the actual private comment in HubSpot. Always allow changes future approval behavior, so do not use it for a one-action test. Workspace billing may include other agents' work.

This separate publishing-support example from eesel's documentation illustrates the private draft destination. It is not a result of the proposed access-request test.

Choose the workflow your team can verify

HubSpot's native agents are worth evaluating when their role, sources, and deployment controls fit your job. Use their built-in tests rather than assuming another product is needed for previewing answers.

For an eesel teammate connected to HubSpot, the CLI gives your technical team a practical way to inspect setup and work with the same agent from a terminal or coding workflow. That is useful when you want repeatable checks without moving the support team out of its inbox.

Try eesel and use the CLI to inspect one HubSpot support workflow. Expand only after the policy answer, permissions, private delivery, and review effort meet your requirements.