# HubSpot Agent Hub Agents in HubSpot for customer service, sales, marketing and data. Agent Hub brings together prebuilt and custom agents that work in the CRM and are billed through credits. Canonical profile: https://ki-agentenkompass.de/en/agents/hubspot-agent-hub/ Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/) Sources reviewed: 2026-11-05 Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests. ## Overview The name "Breeze Agents" is no longer the current umbrella term. HubSpot now lists the agents under Agent Hub, and Breeze Studio is now called agent builder. Breeze Assistant remains as the assistant that, according to HubSpot, puts matching agents to work. The older label still appears in earlier announcements. - Provider: HubSpot - Product family: HubSpot Agent Hub - Profile type: Agent workspace in a CRM platform - Environment: HubSpot web interface with Agent Hub and agent inbox; also started through CRM cards, automations and Breeze Assistant - Connections: HubSpot CRM and HubSpot actions; external systems through MCP servers; ZoomInfo, Apollo, Surfe and Seamless for the prospecting agent - Setup: A Super Admin enables the AI settings. You then activate agents in Agent Hub or build your own in the agent builder. For the customer agent you also connect channels and content. - Pricing model: Credits at $0.010 each. Customer agent $0.50 per resolution, prospecting agent $1 per lead. Agent builder shows the button "Buy now for $7/mo"; custom agents also cost credits. ## Key features ### Agent Hub Shows all agents and agentic workflows in the account. According to the article, you activate, configure and monitor them in one place. Source: https://knowledge.hubspot.com/ai/understand-agent-hub ### Customer Agent Answers questions through chat, email, forms, WhatsApp, Facebook Messenger and phone and hands difficult cases to your team with context. Source: https://www.hubspot.com/products/artificial-intelligence/ai-customer-service-agent ### Prospecting Agent Researches accounts, finds contacts and drafts outreach. You can review drafts beforehand or later let it send without review. Source: https://www.hubspot.com/products/sales/ai-prospecting-agent ### Agent Builder Builds custom agents with inputs, instructions, tools and knowledge. CRM write access is not on by default. Test runs use no credits. Source: https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder ### Agent Inbox Lists runs with status, trigger and estimated credits. You see sources and thinking steps, refine results in chat and export them. Source: https://knowledge.hubspot.com/ai/review-agent-output ### Context and knowledge vaults The context tab holds brand and ideal customers. Up to 50 vaults combine files, HubSpot content and segments for projects and agents. Source: https://knowledge.hubspot.com/ai/manage-breeze-context-with-knowledge-vaults ### Agent Marketplace Lists prebuilt HubSpot agents. Custom agents from the builder cannot be listed there. Five agents ended for new installations on 23 July 2026. Source: https://knowledge.hubspot.com/ai/install-agents-from-the-agent-marketplace ## Example workflows ### Answer support chat - Your input: Knowledge base, website and CRM data as sources and a connected chat channel. - Possible result: The customer agent answers questions with sources and hands difficult cases to your team. - What to check: Test before launch, roll out gradually and watch the resolution rate and credit consumption. ### Reach out to contacts - Your input: A target customer list in the CRM and the prospecting agent in review mode. - Possible result: Researched contacts and personalized drafts per lead. - What to check: Check each draft before sending and plan credits per recommended lead. ## Requirements and limits ### Edition and credits Requirement: Agent Hub exists for Starter, Professional and Enterprise. Some HubSpot agents and agentic workflows need Professional or Enterprise. Credits are required. Source: https://knowledge.hubspot.com/ai/understand-agent-hub ### Super Admin for AI settings Requirement: Only Super Admins manage AI settings. Files data is off by default, while CRM and conversation data are on. Source: https://knowledge.hubspot.com/account-management/manage-your-ai-settings ### Credits expire monthly Limitation: Unused credits expire at the end of the month. Source: https://www.hubspot.com/products/artificial-intelligence/credits ### Resolution counts after 72 hours Limitation: A customer agent resolution counts if no handover to a human follows within 72 hours. A lead qualification can also count as a resolution. Source: https://www.hubspot.com/products/artificial-intelligence/ai-customer-service-agent ### Contradictions on the pages Not yet confirmed: The prospecting page names Starter to Enterprise, the April announcement Pro and Enterprise. The announcement names 28 trial days, the knowledge base 14 days for the customer agent. Source: https://www.hubspot.com/company-news/hubspots-customer-agent-and-prospecting-agent-now-you-pay-when-the-task-is-complete ### Retention of runs Not yet confirmed: The agent inbox article gives no retention period for runs and results. Source: https://knowledge.hubspot.com/ai/review-agent-output ## A good fit if … - your team already works with HubSpot CRM - you want to bill customer service or outreach by outcome through credits - you want to build custom agents without code on CRM data ## Less suitable if … - you do not use HubSpot and are looking for a standalone agent - you need predictable fixed costs instead of consumption - you are looking for a product with model training off by default Editorial assessment based on provider documentation. Not a product test. ## Usable AI models Fixed choice across providers HubSpot's model card lists models from Anthropic, OpenAI and Google for agents; HubSpot selects these providers itself. The help for the agent builder shows no model choice and does not say which model each agent uses. Your own keys are available for the Custom LLM workflow action on Enterprise plans, but this is not documented for agents. - Model families named: Anthropic (Claude), OpenAI (GPT), Google (Gemini) - Model choice: Not stated - Your own API key: Not stated - Router or custom endpoint: Not stated - Local models: Not stated Partly documented · Sources accessed on: 2026-11-05 - [HubSpot: AI model cards](https://trust.hubspot.com/ai) - [HubSpot: AI cloud infrastructure FAQ](https://knowledge.hubspot.com/hubspot-ai-cloud-infrastructure-frequently-asked-questions) - [HubSpot: Agent builder](https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder) - [HubSpot: Custom LLM workflow actions](https://knowledge.hubspot.com/integrations/create-and-use-custom-llm-workflow-actions) ## Europe & regional availability Hosting article, AI infrastructure FAQ, sub-processor page and model training article reviewed. They name no regional restriction for individual agents, but this alone does not establish availability. Reviewed scope: Hosted CRM with EU data center in Germany Reviewed on: 2026-11-05 ### EU hosting in Germany With requirements Accounts in the EU data center are hosted in Germany. The region follows from the IP at sign-up; paid accounts can migrate. Processing elsewhere is possible in named exceptions such as sub-processors and support. Region covered: EU Documented reason: Applies depending on account region and with the named exceptions. Official source: https://knowledge.hubspot.com/account-security/hubspot-cloud-infrastructure-and-data-hosting-frequently-asked-questions ### AI processing for EU accounts With requirements The sub-processor page names OpenAI with location EEA and Switzerland for EU data center customers. For web search results in HubSpot AI (Bright Data) a US location is also listed in that column. According to the FAQ, HubSpot forbids providers from training on customer data. Region covered: EU/EEA and Switzerland Documented reason: Differs by provider and feature; the list can change. Official source: https://legal.hubspot.com/sub-processors-page Official source: https://knowledge.hubspot.com/hubspot-ai-cloud-infrastructure-frequently-asked-questions ### Model training on by default Availability still unclear By default HubSpot may use customer data for its own AI models, including across accounts. Super Admins can turn this off; data already used stays in the models. The article names no regional differences. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The article names no region and no legal basis. Official source: https://knowledge.hubspot.com/account-management/hubspot-ai-mode-training ### United Kingdom and Switzerland Availability still unclear The hosting article names no separate hosting region for either. Switzerland appears only in connection with the Data Privacy Framework. Region covered: United Kingdom and Switzerland Documented reason: No separate region; the page does not state conditions for such customers. Official source: https://knowledge.hubspot.com/account-security/hubspot-cloud-infrastructure-and-data-hosting-frequently-asked-questions ## Accounts & permissions Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use. Configuration covered: Hosted CRM with roles and AI settings Sources reviewed: 2026-11-05 ### Account used Documented: Allowed are password, social login, passkeys and SSO; SSO only in Professional and Enterprise. Super Admins set the sign-in methods. Evidence: https://knowledge.hubspot.com/account-security/restrict-which-login-methods-users-can-use-to-access-your-account ### Admin approval Documented: Runs can wait for tool approval in the agent inbox. With the prospecting agent you choose review or autonomous sending. With the customer agent, handover rules take this role. Evidence: https://knowledge.hubspot.com/ai/review-agent-output Evidence: https://www.hubspot.com/products/sales/ai-prospecting-agent Evidence: https://knowledge.hubspot.com/customer-agent/understand-the-customer-agent ### Data access & team sharing Documented: Agents read CRM data and the web, write access must be selected separately, and external systems come through MCP. Runs from automations use the agent owner's permissions. Evidence: https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder Evidence: https://knowledge.hubspot.com/ai/review-agent-output ## Prices and included usage Pricing checked: 2026-11-05 Public credit prices; Agent Hub is included for paid editions, and the agents mostly bill through credits. ### HubSpot Credits $0.010 per credit; packs at $10 per 1,000 credits Agent consumption. Included credits vary by product and tier; unused ones expire at the end of the month. Price source: https://www.hubspot.com/products/artificial-intelligence/credits ### Customer Agent $0.50 per messaging resolution (50 credits); $0.50 per voice minute (50 credits) According to the page, from Professional. Only resolved conversations are billed. Price source: https://www.hubspot.com/products/artificial-intelligence/ai-customer-service-agent ### Prospecting Agent $1 per lead recommended for outreach (100 credits) Research, contacts and drafts. Edition statements contradict each other. Price source: https://www.hubspot.com/products/sales/ai-prospecting-agent ### Agent Builder Button "Buy now for $7/mo"; the page does not state whether this is per seat or account; runs additionally in credits Included in Starter, Professional and Enterprise; test runs without credits. Price source: https://www.hubspot.com/products/artificial-intelligence/ai-agents Further rates on the credits page: content agent 1,000 credits per piece, nurture agent 10 per email, data agent 10 per response, revenue agent 500 per invoice. Third-party pricing was not used as evidence. Pricing details: https://www.hubspot.com/products/artificial-intelligence/credits ## Operation & oversight Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured. Deployment covered: Hosted operation by HubSpot Sources reviewed: 2026-11-05 ### Files and working state Documented with conditions: Knowledge lives in vaults and context and stays available across runs. Runs are listed in the agent inbox, and the article gives no retention period. Evidence: https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder Evidence: https://knowledge.hubspot.com/ai/manage-breeze-context-with-knowledge-vaults Evidence: https://knowledge.hubspot.com/ai/review-agent-output ### Sign-ins Documented with conditions: SSO in Professional and Enterprise, plus password, social login and passkeys. The builder article does not say how MCP credentials are managed. Evidence: https://knowledge.hubspot.com/account-security/restrict-which-login-methods-users-can-use-to-access-your-account Evidence: https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder ### Starting further runs Documented: Agents start from the inbox, CRM cards, Breeze Assistant or an automation. According to the page, event-based triggers exist in Professional and Enterprise. Evidence: https://knowledge.hubspot.com/ai/review-agent-output Evidence: https://www.hubspot.com/products/artificial-intelligence/ai-agents ### Human intervention Documented: The "Needs approval" status waits for tool approval. According to the page, you decide which actions run alone and set guardrails. The customer agent hands over to humans at low confidence. Evidence: https://knowledge.hubspot.com/ai/review-agent-output Evidence: https://www.hubspot.com/products/artificial-intelligence/ai-agents Evidence: https://knowledge.hubspot.com/customer-agent/understand-the-customer-agent ### Detecting errors Documented with conditions: The inbox shows the statuses Success, Running, Failed and Waiting and error notes such as "Missing tool". According to the settings, the account audit log covers changes to model training; the article names no log per agent action. Evidence: https://knowledge.hubspot.com/ai/review-agent-output Evidence: https://knowledge.hubspot.com/account-management/manage-your-ai-settings ### Resuming after failures Documented with conditions: You can have results revised in chat, and the agent then runs again. The article does not describe resuming at the point of failure. Without credits, new customer agent conversations follow the handover rule. Evidence: https://knowledge.hubspot.com/ai/review-agent-output Evidence: https://knowledge.hubspot.com/customer-agent/understand-the-customer-agent ### Editorial assessment HubSpot documents permissions, statuses and approvals in its knowledge base. The reviewed pages say nothing on retention, resuming or a log per action; costs depend on consumption. ## Memory and conversation continuity Scope: Memories in Breeze Assistant; for agents separately context and knowledge vaults Sources accessed: 2026-11-05 Breeze Assistant stores preferences as memories that you can delete. Agents use vaults and context; the articles describe no separate agent memory across runs. ### Storage and scope Partly documented: Breeze Assistant stores name, role, tone and recurring instructions as memories. Agents use knowledge vaults and context. The article does not say where the data is stored. - [HubSpot: Use Breeze Assistant](https://knowledge.hubspot.com/ai/use-breeze-assistant) - [HubSpot: Knowledge vaults](https://knowledge.hubspot.com/ai/manage-breeze-context-with-knowledge-vaults) ### Selection and timing Partly documented: You ask Breeze Assistant in chat to remember or forget something. The article names no automatic selection rules. - [HubSpot: Use Breeze Assistant](https://knowledge.hubspot.com/ai/use-breeze-assistant) ### Inspect, edit and delete Documented: Under More and Memories you view and delete entries. For agents you delete vault content in the vault; you can change memories through chat. - [HubSpot: Use Breeze Assistant](https://knowledge.hubspot.com/ai/use-breeze-assistant) - [HubSpot: Knowledge vaults](https://knowledge.hubspot.com/ai/manage-breeze-context-with-knowledge-vaults) ### Agent and channel boundaries Partly documented: Memories belong to the user, and project content applies within the project. The article does not say whether builder agents read memories. - [HubSpot: Understand Breeze Assistant](https://knowledge.hubspot.com/ai/understand-breeze-assistant) ### Retrieval Documented: Breeze Assistant can search earlier conversations; conversation history is on by default. Knowledge from vaults stays available to agents across runs. - [HubSpot: Use Breeze Assistant](https://knowledge.hubspot.com/ai/use-breeze-assistant) - [HubSpot: Create and customize agents](https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder) ### Session reset and compaction Not publicly established: The reviewed articles do not mention reset or compaction. The customer agent article does not say whether it retains information across conversations. - [HubSpot: Understand the customer agent](https://knowledge.hubspot.com/customer-agent/understand-the-customer-agent) ## Context A HubSpot account with a matching edition and credits, several agents only from Professional. AI model training is on by default. The reviewed pages do not cover resuming aborted runs or retention of runs. HubSpot product and pricing pages, knowledge base articles, announcements and the sub-processor page reviewed. The pages contradict each other on trial periods and on the edition for the prospecting agent. Example tasks are editorial. Hands-on tests are still pending. ## Open points Europe, availability still unclear: - Model training on by default - United Kingdom and Switzerland Requirements and limits, not yet confirmed: - Contradictions on the pages - Retention of runs ## Sources - https://www.hubspot.com/products/artificial-intelligence - https://www.hubspot.com/products/artificial-intelligence/ai-agents - https://www.hubspot.com/products/artificial-intelligence/breeze-ai-agents - https://knowledge.hubspot.com/ai/understand-agent-hub - https://knowledge.hubspot.com/ai/create-and-customize-agents-in-the-agent-builder - https://knowledge.hubspot.com/ai/review-agent-output - https://www.hubspot.com/products/artificial-intelligence/credits - https://www.hubspot.com/products/artificial-intelligence/ai-customer-service-agent - https://knowledge.hubspot.com/customer-agent/understand-the-customer-agent - https://www.hubspot.com/products/sales/ai-prospecting-agent - https://www.hubspot.com/company-news/hubspots-customer-agent-and-prospecting-agent-now-you-pay-when-the-task-is-complete - https://www.hubspot.com/company-news/fall-26-spotlight - https://www.hubspot.com/company-news/build-your-ai-team - https://knowledge.hubspot.com/account-management/manage-your-ai-settings - https://knowledge.hubspot.com/account-management/hubspot-ai-mode-training - https://knowledge.hubspot.com/hubspot-ai-cloud-infrastructure-frequently-asked-questions - https://knowledge.hubspot.com/account-security/hubspot-cloud-infrastructure-and-data-hosting-frequently-asked-questions - https://legal.hubspot.com/sub-processors-page - https://knowledge.hubspot.com/account-security/restrict-which-login-methods-users-can-use-to-access-your-account - https://knowledge.hubspot.com/ai/manage-breeze-context-with-knowledge-vaults - https://knowledge.hubspot.com/ai/use-breeze-assistant - https://knowledge.hubspot.com/ai/understand-breeze-assistant - https://knowledge.hubspot.com/ai/install-agents-from-the-agent-marketplace ## Provider HubSpot offers a CRM platform for marketing, sales and service. - Operator: HubSpot - Organization: Provider - Website: https://www.hubspot.com - Reviewed: 2026-11-05 - Provider source: https://www.hubspot.com/products/artificial-intelligence [Sources and method](https://ki-agentenkompass.de/en/sources/)