Two of the most talked-about agent launches of the past fortnight look like competitors. They are not, and confusing them is how an enterprise ends up piloting the wrong thing.
By the end of this page you will know what OpenAI Dots and Claude Managed Agents actually are, what each costs in real numbers, whether a Hong Kong organisation can use either one directly, where each wins, and which buyer should choose which. We also cover a third route for teams that need an agent working in Hong Kong this quarter, and we are explicit about its limits too.
What is the difference between OpenAI Dots and Claude Managed Agents?
OpenAI Dots is a ready-made, always-on agent that staff use inside ChatGPT, Slack or Teams, bundled into certain ChatGPT plans. Claude Managed Agents is a developer runtime: a set of APIs for building and hosting your own agents on Anthropic's infrastructure, billed by usage. One is bought per seat; the other is built per use case.
OpenAI announced Dots at DevDay on 29 September 2026. According to OpenAI’s launch post, each dot runs on GPT-6 Astra, has its own cloud computer, connects to more than 4,000 apps through plugins and can keep working on a goal around the clock. Specialist dots for organisations get their own identity and are managed through Microsoft Agent 365.
Claude Managed Agents launched in April 2026. Your developers define an agent, its tools and its guardrails; Anthropic runs the execution environment, session state, permissions and tracing. It is the infrastructure underneath an agent, not the agent your staff talk to.
So the real buying question is not "which agent is smarter". It is whether you want to buy an agent for employees or build agents into your own processes and products.
How much do OpenAI Dots and Claude Managed Agents cost?
Dots carry no separate price: the first dot is included in ChatGPT Pro plans at US$100, US$200 or US$500 a month, or a Business Premium seat. Claude Managed Agents costs standard Claude API token rates plus US$0.08 per active session-hour, with idle time not billed and no seat fee.
OpenAI Dots pricing
--- First dot included in Pro 100, Pro 200, Pro 500 or Business Premium, per DevDay pricing coverage.
--- Conversations with a dot do not count toward ChatGPT usage limits.
--- Enterprise, Edu and Healthcare: beta, switched off by default until an admin enables it.
--- Additional dots and higher monthly work allowances: not yet priced.
Claude Managed Agents pricing
--- Tokens at standard Claude API rates for the chosen model.
--- US$0.08 per session-hour while the session is running, metered to the millisecond.
--- Batch, fast-mode and data-residency discounts do not apply, according to TrueFoundry’s pricing breakdown.
--- Web search billed separately at US$10 per 1,000 searches.
A worked example from Anthropic's documentation, as cited by TrueFoundry: a one-hour session on Claude Opus 5 using 50,000 input and 15,000 output tokens costs about US$0.71, of which only US$0.08 is runtime. At fleet scale the runtime line grows with agent count, which is why usage-based budgeting matters.
If your finance team is new to metered AI, UD's guide to usage-based AI billing covers how to set caps before the first invoice.
Can Hong Kong enterprises use OpenAI Dots or Claude Managed Agents directly?
Not straightforwardly. Hong Kong is not on OpenAI's or Anthropic's supported-region lists, Dots is offered only in OpenAI's eligible markets, and Claude Managed Agents' hosted runtime is US-based during its beta. Using a VPN to get around regional checks likely breaches both providers' terms and is not a basis for enterprise deployment.
Regional trackers such as Digital in Asia’s LLM access tracker list Hong Kong as unsupported for direct ChatGPT and Claude accounts. Even large firms are affected: Bloomberg reported in April 2026 that Goldman Sachs staff in Hong Kong lost access to Claude.
Indirect routes exist. OpenAI models reach Hong Kong organisations through Microsoft 365 Copilot and Azure, and specialist dots are being integrated with Microsoft Agent 365. Claude models are offered through some cloud marketplaces and inside Copilot where the tenant enables them. Each route has its own regional availability, so confirm it with your Microsoft or cloud account team in writing.
For regulated firms, availability is only half the issue. Data location and PDPO obligations still sit with you, whichever route you take.
Is OpenAI Dots better than Claude Managed Agents for enterprise work?
Neither is better across the board, because they solve different problems. Dots wins on time-to-value for individual knowledge workers already on ChatGPT. Claude Managed Agents wins when you need custom, auditable agents embedded in your own workflows. Both lose on direct Hong Kong availability and local data residency.
Same criteria, both options
--- What you get: Dots, a finished personal agent; Managed Agents, a hosted runtime you build on.
--- Who sets it up: Dots, the end user or IT admin; Managed Agents, your developers or a partner.
--- Pricing model: Dots, bundled into seats; Managed Agents, tokens plus US$0.08 per active hour.
--- Model choice: Dots, OpenAI models only; Managed Agents, Claude models only.
--- Governance: Dots, specialist dots via Microsoft Agent 365; Managed Agents, developer-defined permissions and tracing.
--- Maturity: Dots rolling out, enterprise in beta; Managed Agents in beta.
--- Direct Hong Kong availability: neither.
What are the limitations of each option, including UD Cloud AI Staff?
Every option here has a real limitation. Dots is personal productivity tooling with unpriced scale-up. Claude Managed Agents demands engineering capacity and locks you to one model family. UD Cloud AI Staff is local and fast to deploy but is not a developer platform and does not publish a monthly price.
--- OpenAI Dots: designed around individual users rather than shared business processes; extra dots unpriced; OpenAI's own safety post says dots can make mistakes and consequential work should be reviewed; not directly available in Hong Kong.
--- Claude Managed Agents: requires developers to design, test and maintain agents; Claude-only; runtime cost compounds across long-running fleets; hosted runtime US-based in beta with no on-premises option.
--- UD Cloud AI Staff: role-based AI employees for functions such as customer service, admin and marketing, not a toolkit for building your own products; no published monthly subscription price, so you will need a quote.
Published UD terms on the campaign page, as checked on 12 October 2026: HK$0 activation fee, UD specialists assisting throughout, an AI employee live in as little as three business days, and a locally deployed service with data kept in Hong Kong. For a build-versus-buy view on Microsoft's tooling, see Copilot Studio vs AI Staff.
Which option should each type of Hong Kong buyer choose?
Choose Dots if your teams already work in ChatGPT through an eligible route and want personal productivity gains. Choose Claude Managed Agents if you have engineers and a specific workflow to automate. Choose a locally deployed AI employee if you need a customer-facing or back-office role running in Hong Kong this quarter.
--- A regional bank or insurer with strict data rules: neither Dots nor Managed Agents directly; start with Microsoft-channel options already approved by risk, or a Hong Kong-hosted AI employee for a contained role.
--- A logistics or property group with an in-house tech team: Claude Managed Agents through an approved cloud route, for a defined workflow such as shipment exception handling.
--- A professional services firm whose partners live in ChatGPT: Dots, once eligibility and the enterprise beta are confirmed for your tenant.
--- A mid-market company with no AI engineers and a service backlog: a managed AI employee with local deployment and a fixed scope.
If you are also weighing productivity-suite agents, UD's comparison of Gemini Agent and Microsoft 365 Copilot covers that adjacent decision.
What should you do before signing up for any agent platform?
Before committing budget, confirm legal availability for your entity in writing, pick one measurable workflow, cap spend for a 60-day pilot, decide who reviews agent output, and define the exit criteria. These five steps prevent the most common failure: a pilot that runs, costs money and proves nothing.
--- Week 1: written confirmation of availability and data handling for your Hong Kong entity.
--- Week 2: one workflow with a baseline, for example average handling time for a specific enquiry type.
--- Weeks 3 to 8: pilot with a hard monthly cap and a named human reviewer.
--- Week 9: compare against the baseline and decide to scale, switch or stop.
Write the exit criteria before the pilot starts, not after. A useful format is three numbers agreed with the business owner: the minimum improvement that justifies scaling, the maximum monthly spend, and the error rate at which the pilot stops automatically. Teams that skip this step usually end up extending a pilot that never had a definition of success, which is how agent budgets drift without anyone approving them.
OpenAI Dots vs Claude Managed Agents: key facts
--- OpenAI Dots launch: 29 September 2026, OpenAI DevDay.
--- Dots price: first dot included in ChatGPT Pro (US$100, US$200 or US$500 a month) or Business Premium; extra dots unpriced.
--- Dots for Enterprise: beta, off by default.
--- Claude Managed Agents launch: April 2026, beta.
--- Managed Agents price: Claude API token rates plus US$0.08 per active session-hour; web search US$10 per 1,000.
--- Direct Hong Kong availability: not supported by either provider.
--- UD Cloud AI Staff: HK$0 activation, live in as little as 3 business days, data kept in Hong Kong; monthly price on quote.
Conclusion: decide whether you are buying an agent or building one
Dots and Claude Managed Agents are both serious products, but they answer different questions. If your aim is personal productivity, Dots is the simpler purchase where it is available. If your aim is a custom, auditable agent inside a business process, Managed Agents is the stronger foundation, provided you have the engineering capacity and an approved access route.
For many Hong Kong enterprises, the deciding constraint is neither capability nor price. It is whether the service can legally and safely run here, with data where your regulators expect it.
We understand the cold edges of AI and the hard parts of your work, and UD has walked with Hong Kong enterprises for twenty-eight years, making technology a partnership with warmth.
Reviewed by the UD enterprise AI team. Prices and availability checked against the linked sources on 13 October 2026. Third-party figures are attributed; confirm current terms with each vendor before purchase.
Get an AI employee working in Hong Kong in three business days
If you need an agent live this quarter without an engineering team or an overseas access route, start with one role. We'll walk you through every step, from choosing the role and setting scope to Hong Kong deployment and performance tracking, with a HK$0 activation fee.
