Agent Hub setup and configuration.
The agents HubSpot gives you, configured against your real pipelines, properties and playbooks rather than left on defaults.
HubSpot's own AI (now Agent Hub, previously Breeze) does more than most teams realise, and much of it sits switched off or running against messy data. We configure it properly, then build custom Claude agents for the work it was never designed to do.
Agent Hub and Breeze are genuinely good at the work they were designed for. They stall when a job needs your own logic, a system outside HubSpot, or data the CRM does not hold. Knowing which is which saves a great deal of money.
HubSpot's agents draft from whatever is in the CRM. Where records are patchy, the output is confidently wrong.
We fix the data foundations first, so the agents answer from records that are complete, deduplicated and current.
Usage-based AI bills by the unit of work. Without governance you pay for drafts and answers nobody uses.
We scope which agents run where, set the guardrails, and report on what the spend actually returns.
The moment a process needs your pricing rules, an ERP lookup or a system HubSpot does not reach, the built-in agents stop.
We build a Claude agent for that step, wired into HubSpot and the systems around it, so the process completes end to end.
Teams quietly stop using AI suggestions after the first few misses, and the subscription keeps renewing.
Role-based rollout, clear guardrails on what to check, and before-and-after measurement of real usage.
HubSpot's agents draft from whatever is in the CRM. Where records are patchy, the output is confidently wrong.
We fix the data foundations first, so the agents answer from records that are complete, deduplicated and current.
Usage-based AI bills by the unit of work. Without governance you pay for drafts and answers nobody uses.
We scope which agents run where, set the guardrails, and report on what the spend actually returns.
The moment a process needs your pricing rules, an ERP lookup or a system HubSpot does not reach, the built-in agents stop.
We build a Claude agent for that step, wired into HubSpot and the systems around it, so the process completes end to end.
Teams quietly stop using AI suggestions after the first few misses, and the subscription keeps renewing.
Role-based rollout, clear guardrails on what to check, and before-and-after measurement of real usage.
Swipe to explore
We are a HubSpot Diamond partner and a software engineering firm, so we can configure HubSpot's own AI properly and write real software where it runs out of road. We work with UK companies running HubSpot, typically from around £3m in revenue upwards.
The agents HubSpot gives you, configured against your real pipelines, properties and playbooks rather than left on defaults.
Deduplication, property hygiene and structure, so the AI answers from records worth trusting. Usually the highest-return work on the list.
Where HubSpot's no-code agent builder can do the job, we build it there and hand it over, rather than selling you engineering you do not need.
For the work HubSpot's AI was not designed to do: your own business logic, systems outside HubSpot, and long multi-step processes.
HubSpot bills its AI by units of work. We scope where agents run, cut the waste and report on what the spend returns.
Role-based training and clear guardrails, so your team actually uses what you are paying for.
We review your portal, your data quality and the AI you already pay for, and pin down which jobs are worth automating.
We fix the data foundations and configure HubSpot's agents against your real pipelines and processes.
Where the platform stops, our engineers build a custom Claude agent and integrate it properly.
We train the team role by role, monitor usage and keep the spend honest.
HubSpot's agents are good at CRM-shaped work. These are the jobs where a custom Claude agent earns its keep.
Order entry, quoting against ERP stock, contract checks: anything needing a system HubSpot does not natively reach. We built an AI order-entry agent for a UK builders' merchant that reads incoming orders and writes them into the system.
Pricing logic, eligibility, risk scoring and approval rules that are specific to your business and will not come out of a generic model.
Research, reconciliation and document workflows that need to run for minutes and complete a job, rather than return a single suggestion.
The most common reason HubSpot AI disappoints is that nobody trusts it enough to use it. We fix that the same way we run HubSpot training: structured, role by role, and measured.
Short, hands-on sessions with practice gaps between them, so habits embed instead of fading after a one-off workshop.
What a salesperson should hand to an agent is not what a service lead should. We train against the real jobs people do.
People learn what to check and where the limits sit, so adoption is safe as well as fast.
A before-and-after competence survey plus usage reporting, so you can see whether the spend moved anything.
Most clients start here, with a fixed-scope diagnostic, before committing to a delivery project.
A fixed-scope diagnostic of your portal: data quality, which AI features you are already paying for, what is worth switching on, and what genuinely needs custom engineering.
We've delivered 300+ HubSpot implementations as a firm, and we specialise in Anthropic's Claude for AI work. Technology only delivers value when people use it, so adoption is built into everything we ship.
Agents and integrations designed, tested and shipped by our own engineers. Nothing is subcontracted.
Transformation projects shipped for B2B teams across the UK and beyond.
We've gone deep on Anthropic's Claude rather than dabbling across every model, so you get engineers who know it inside out.
<p>Largely, yes. HubSpot renamed Breeze Agents to Agent Hub during 2026, and Breeze studio became Agent Builder. You will still see the Breeze name in places, including parts of HubSpot's own documentation and product areas such as Breeze Projects. If you bought Breeze, you have Agent Hub.</p>
<p>Often not, and we will say so. If the job is CRM-shaped, HubSpot's own agents or Agent Builder will usually handle it, and configuring those properly is cheaper than commissioning software. Custom Claude earns its place when the process leaves HubSpot, depends on your own business rules, or needs to run as a long multi-step job.</p>
<p>HubSpot has moved to usage-based pricing: you pay per unit of work, such as a drafted lead, a resolved conversation or an answered question, and custom agents consume HubSpot credits. The agent features sit on Professional and Enterprise subscriptions. Rates change, so check HubSpot's current pricing rather than relying on a figure here. Our job is to make sure you are not paying for work nobody uses.</p>
<p>The agent features sit on Professional and Enterprise tiers. If you are on Starter, we will tell you what upgrading would actually unlock for your use case before you spend anything on it.</p>
<p>Usually that is the first job. HubSpot's AI answers from your CRM records, so deduplication, property hygiene and structure typically return more than any amount of agent configuration. We treat it as part of the work rather than an upsell.</p>
<p>We are a HubSpot Diamond Solutions Partner, accredited for Custom Integration and Onboarding. Where Claude is concerned we are independent: we are not an Anthropic reseller and we do not sell Claude licences, so you buy Claude directly from Anthropic. What we provide is the engineering.</p>
Read more on HubSpot AI: AI for HubSpot: the full guide · Breeze vs custom Claude agents · HubSpot Breeze Customer Agent · HubSpot Breeze Prospecting Agent · Breeze Assistant · Breeze pricing · HubSpot AI agents compared
Not on HubSpot, or want a custom Claude agent regardless of CRM? See Claude implementation and custom AI agents.
SpotDev is a HubSpot Diamond Solutions Partner. SpotDev is an independent consultancy and is not affiliated with or endorsed by Anthropic. Claude is a trademark of Anthropic, PBC. HubSpot and Breeze are trademarks of HubSpot, Inc.