A practical guide to the HubSpot AI agent marketplace

A marketplace card can tell you what an AI product promises. It cannot show whether the installed agent will use the right account, read the right policy, or leave its draft where your team expects it. I would treat those as separate checks.

For HubSpot, there is another distinction to make first: its Agent Marketplace and its broader app marketplace do not contain the same kind of product. This guide explains the difference and shows how eesel CLI can help evaluate a connected support teammate.

Understand what you are installing

The HubSpot Marketplace separates Apps, Agents, Workflows, and other categories. A connector can provide access to data without being a ready-to-run support agent. A workflow template can describe a process without supplying every required integration or permission.

The label “AI” does not answer these questions. Nor does a high install count establish that an app suits your support queue or privacy requirements.

Add a native HubSpot agent

HubSpot's installation guide describes Agent Marketplace as a beta containing HubSpot's prebuilt agents. Custom Agent Builder agents are not eligible for listing there.

Open More if it appears in your navigation, then use Agents > Agent Marketplace. Select Explore all, open a card, and inspect its creator, overview, functionality, pricing, and support information before clicking Add. Then adapt its instructions and context to your business.

The guide requires Super Admin or Agent Builder permission for creating and customizing agents, along with the relevant AI settings and task-specific permissions. Verify the requirements in your account.

Availability also changes. HubSpot's guide says Social Post Agent and several other agents stopped accepting new installations on July 23, 2026, though existing installations remain. Do not use an older screenshot as proof that an agent can still be added.

Evaluate a partner app on its own terms

The eesel app listing is titled AI agents for customer service and identifies eesel AI as its builder. It belongs to the partner-app route, not the native template installation above.

Before approving a support app, establish the exact portal and the data it will receive. Read access can expose sensitive information even if every write action is turned off. Review permissions, data handling, support ownership, and the path for stopping automation.

A useful first job is narrow: draft an answer to an approved support question for a human to review. Do not start by authorizing every available action simply because it appears in the listing.

For eesel, the current HubSpot guide explains the actual connection and trigger behavior. Use that operational documentation to test the claims that matter to your team.

Check the bill separately from the listing

Native HubSpot features may consume HubSpot Credits. Check the allowance, the feature's chargeable unit, and the account's limits. If you buy extra capacity, review whether auto-upgrades or overages apply.

Partner-app billing is separate. The eesel listing still displays older Team and Business plan language, while the current eesel pricing describes pay-as-you-go tasks. Confirm current terms before approving a purchase; do not treat a marketplace display price as a contract.

Record a baseline before the pilot. Compare actual usage and review time after the test, rather than assuming the cheapest displayed plan produces the cheapest completed work.

Inspect the installed teammate through eesel CLI

The eesel CLI gives a person, script, or coding agent a terminal interface to the same eesel teammate used in the dashboard. Its JSON output lets Claude Code, Cursor, or Codex inspect the connection, source readiness, instructions, and runs.

This is not a command-line interface for adding HubSpot's native marketplace agents. It operates eesel and its connected HubSpot support workflow.

Confirm the account before running a test

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. Agent selection is not isolation: the chosen agent still has its actual integrations and permissions. Review existing automations and writes before connecting a production portal.

If HubSpot is absent and you have approved the data access:

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

A person with App Marketplace Access must select the correct account and approve the browser authorization. A coding agent can prepare the steps, but it should hand the authorization link to that person and wait.

The authorization screen in eesel's documentation identifies the partner app. Expand and review the requested access before connecting your account.

Check what connection has imported

The HubSpot connection starts indexing recently closed tickets and copying permitted CRM records. Make that data-handling decision before authorization, not after running a sample question.

HubSpot knowledge-base articles are read natively; they need not be public. For copied Deals, Contacts, and Companies, the account's granted permissions and the selected agent's knowledge toggles are separate controls. Review live lookup actions and requester access separately too.

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

This example shows copied CRM records with their search toggles off. It also shows zero knowledge-base documents, illustrating why a connected status alone is not enough.

For a support pilot, confirm that the relevant policy was actually imported. Do not enable your entire contact database to answer a question that only needs a help article.

Rehearse a policy-sensitive answer

Before these fictional chats, disable customer sending and ticket-changing actions, including internal comments. The commands do not change those permissions.

npx @eesel/cli new --name "hubspot-app-policy-check" --agent PILOT_AGENT_ID
npx @eesel/cli chat "Fictional support request: a new employee wants access to a shared workspace and asks support to make them an administrator. Use the approved access-request policy to draft a response for review. Explain what approval or identity checks are needed. Do not inspect real customer records, grant access, create a ticket, post a comment, or send a message." --agent PILOT_AGENT_ID
npx @eesel/cli chat "The employee says their manager approved it verbally, but no approval record is provided. Revise the answer without claiming that approval has been verified or that administrator access was granted." --agent PILOT_AGENT_ID

The follow-up continues the same conversation. Check the cited policy, the missing evidence, and whether the proposed reply preserves the distinction between a reported approval and a verified one.

CLI chats are billed work. They test the written answer, not whether HubSpot events arrive or whether a comment lands on the right conversation. The CLI's --dry-run flag previews a server call; it is not a quality simulation.

Verify one real integration event

Choose a test inbox conversation with fictional customer details. Configure a narrowly filtered automation for Call @eesel in an inbox comment. Review its scope before requesting creation, because chat-created automations are enabled when created.

Keep Reply to customer off, require approval for Add internal comment, and leave unrelated writes disabled. Then add a fresh @eesel comment on that inbox conversation. A note on a ticket record does not produce the same event.

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

Review the held comment, approve that one action, and verify the private result in HubSpot. Do not select Always allow unless you intend to change future approval behavior.

An unrelated publishing-support example from eesel's documentation shows the internal-comment destination. It is not a result of the proposed access-request test.

Decide whether the app is ready for wider use

A successful installation is the beginning of evaluation. The outcome that matters is work your team can trust and find in the right place.

Try eesel to evaluate a HubSpot support teammate through the CLI. Use the marketplace to discover it, current docs to configure it, and an inspected pilot to decide what it may do next.