UiPath Pricing 2026: UiPath Cost, License Cost and Agent Unit Rates

Short answer: UiPath publishes two dollar figures and one unit rate table. Automation Cloud Community is free and Basic starts at 25 dollars a month; everything above that is set by sales. Agent work is metered in Agent Units at 2 units per LLM call on the Premium model tier, 1 on Standard and 0.8 on Basic, or 1 unit per agent run if you bring your own model. UiPath does not publish a dollar value for an Agent Unit, so your real cost per agent is fixed in your order form.

If you are building the UiPath business case this quarter, the hard part is not finding the plan names. It is that the unit rate and the dollar rate are published in two different places, and only one of them is published at all. This breaks down what UiPath actually states, what only appears in third-party rate-card write-ups, how to estimate a pool before you have a quote, and the cost line most UiPath agent budgets leave out entirely.

What UiPath publishes on its pricing page

UiPath's public pricing page covers two products, Automation Cloud and Test Cloud. Only the entry tiers carry a number.

• Community, free forever. Build and orchestrate workflows including Maestro. UiPath describes it as ready for production rather than prototyping only. Support is the community forum.
• Basic, from 25 dollars a month. Limited scale on users and robots, European region only, 99.9 percent uptime, Bronze support.
• Standard, contact sales. No scale limits, multi-region or on-premises deployment, and this is the lowest tier that includes agents along with document classification and extraction.
• Enterprise, contact sales. Adds self-healing UI automation, process simulation, bring-your-own-model for AI products, customer-managed encryption keys and credential vaults.

One thing worth noticing before you plan around Basic: agents are a Standard-tier capability. The 25 dollar figure is real but it is not the number for an agentic automation program.

Test Cloud is the one place UiPath prints enterprise-sized numbers, and they are useful as an anchor for how the company sizes add-ons: Performance Testing at 20,000 dollars per 2,500 virtual users, Screenplay at 10,000 dollars per 50,000 runs, and Self-Healing at 5,000 dollars per 5,000 heals. Screenplay is included at no extra cost on Test Cloud Enterprise.

UiPath Agent Unit rates: the one rate table UiPath does publish

This is the table that actually drives agent cost, and it is in UiPath's agent licensing documentation rather than on the pricing page. Agent Units are consumed at both design time and run time, which matters more than it sounds: debug runs and evaluations draw down the same pool as production.

The distinction in the last column is the single most important line in UiPath agent pricing, and it is easy to skim past.

With a UiPath-hosted model, you are billed per LLM call. An agent that reasons in one shot costs one call. The same agent, given a harder task, might loop through twelve tool calls and twelve model calls before it finishes. On the Premium tier that is 24 Agent Units instead of 2, for the same task definition, decided at run time by the model.

With a customer-managed model, you are billed once per agent run, no matter how many times the model is called inside it. You pay your own inference bill to the provider instead, but the UiPath meter becomes flat and predictable. For any agent that does real multi-step work, bring-your-own-model changes the shape of the cost curve completely, not just its level. That capability sits on the Enterprise tier.

How the monthly pool works, and the rule that catches people out

Enterprise customers get a monthly AI and Agentic usage pool that is shared across UiPath's AI and agentic products. Two published mechanics decide whether that pool is usable.

First, the monthly limit is assigned individually to each named user and is not poolable at tenant or organization level. That phrasing is UiPath's own, and it is the constraint most capacity plans get wrong. You cannot size one shared allowance across the team and let the heaviest agent builder draw most of it. Each named user has their own ceiling.

Second, by default each tenant is allocated zero agent units, and consumption draws from the organization pool. So a new tenant does not fail closed with a helpful error about licensing until someone allocates to it, and a production tenant quietly consuming from the org pool is the normal state rather than a misconfiguration. Decide deliberately which tenants get their own allocation.

What UiPath does not publish, and how to read third-party numbers

UiPath publishes rates in units. It does not publish a dollar value for an Agent Unit or a Platform Unit anywhere public. That gap is real, and it is why every "UiPath cost" article you find carries different figures.

Consultancies that work on UiPath renewals do circulate ranges from rate cards they have seen, including claims of 1 to 750 credits per action depending on the capability, API usage metered around 60 units per 10,000 calls, and overage billing beginning in 2027. We are not going to restate those as facts, because we have not seen the rate card and neither has anyone publishing them openly. Treat them as a starting hypothesis for a negotiation, and get the number confirmed in your order form before it goes into a business case.

There is also a second commercial framework worth naming so you ask for the right one. UiPath offers credit-based Unified pricing, which consolidates consumables into a single Platform Unit, alongside license-based Flex pricing. The SDK exposes both meters separately, as Agent Units and Platform Units, so when you ask for a quote, ask which of the two you are being quoted on and which meter each agent draws from. A quote in one framework is not comparable to a quote in the other.

How to size an Agent Unit pool before you have a quote

You can build a defensible estimate without a dollar rate, because the unit side is fully published. Work in units first, then multiply once at the end when sales gives you the rate.

1. Count model calls, not agent runs. For each agent, estimate the average number of LLM calls per completed task. The honest way to get this is to run the agent in a debug environment and read the actual consumption, since design-time execution is metered too.
2. Multiply by the tier rate. Premium 2.0, Standard 1.0, Basic 0.8 per call. Route the agents that do routine extraction or classification to Basic and keep Premium for the ones that genuinely need it. The gap between 0.8 and 2.0 is a 2.5x difference on identical work.
3. Add design-time. Development and evaluation runs consume from the same pool. Teams routinely underestimate this during the build quarter, when debug runs outnumber production runs.
4. Model the tail, not the average. Because hosted-model billing is per call, your worst case is an agent that loops. Estimate a p95 call count as well as a mean, and set max_tool_calls_per_session and max_llm_calls_per_session in UiPath's policy engine so the tail has a hard stop.
5. Check per-user ceilings against your build team. Since the monthly allowance is per named user and not poolable, a team with three heavy builders and ten light ones needs the allocation modeled per person.
6. Price the customer-managed alternative in parallel. Run the same volume at 1 unit per agent run plus your own inference cost. For multi-step agents this is often the cheaper and far more predictable column.

The cost line that never reaches a UiPath invoice

Every number above prices what the agent costs to run. There is a second cost that a UiPath budget never sees: what the agent buys.

We measured this directly rather than assuming it. On 23 September 2026 we downloaded both SDKs UiPath publishes openly, @uipath/uipath-typescript 1.7.2 from npm and uipath 2.14.25 from PyPI, and scanned every property name in them: 1,157 distinct names in the TypeScript SDK and 673 in the Python SDK. A money regex covering amount, currency, budget, spend, price, cost, payment, invoice, dollar and purchase returned three hits in one and two in the other, and every single one is a false positive on a substring, such as feedbackRating matching the letters "fee". The same regex on the Stripe SDK matches 376 of 2,493 property names, so it is certainly not blind.

UiPath's policy engine does ship a check named budget. Its three settings are max_tool_calls_per_session, max_tool_calls_per_minute and max_consecutive_tool_calls. All three count calls. Given that UiPath bills hosted models per call, capping calls is a rational way to cap the UiPath invoice, and it is a genuinely useful control. It just is not a budget in dollars, and the full measurement is in our breakdown of UiPath agentic automation spend controls.

So think of an agent as having two ledgers. The first is Agent Units and Platform Units, metered precisely and capped by call count. The second is the purchase order it submits, the SaaS renewal it approves, the supplier it pays, the flight it books, the API credits it tops up. A procurement agent capped at 50 tool calls per session can still commit 40,000 dollars on call 12, because the amount lives inside the tool arguments and UiPath's only numeric operator is bound to the five per-session call counters.

Budget both ledgers or you have budgeted half the program. For the second one, the controls that work are the ordinary ones finance already understands: a funded wallet or virtual card per agent with a limit in dollars that accumulates across sessions rather than resetting, a merchant and category allowlist, and an approval threshold that fires on the amount. Teams that already map their controls to a compliance framework will recognize the pattern, since UiPath's own policy packs carry a standard name and an OWASP reference field for exactly that reason.

Frequently asked questions

How much does UiPath cost per month?

UiPath Automation Cloud Community is free and Basic starts at 25 dollars a month with limited scale and European hosting only. Standard and Enterprise, which are the tiers that include AI agents, are quoted by sales rather than published. Agent work is billed separately in Agent Units on top of the plan.

How much does a UiPath license cost?

UiPath does not publish per-license prices above the 25 dollar Basic tier. Cost depends on which commercial framework you are on, credit-based Unified pricing or license-based Flex pricing, plus your robot and user counts, term length and negotiated volume. The only published enterprise figures are Test Cloud add-ons, such as 20,000 dollars per 2,500 virtual users for performance testing.

What is a UiPath Agent Unit?

An Agent Unit is UiPath's consumption meter for agent activity, drawn down at both design time and run time. With UiPath-hosted models it is charged per LLM call at 2 units on Premium, 1 on Standard and 0.8 on Basic. With a customer-managed model it is charged once per agent run regardless of how many LLM calls happen inside.

How much does a UiPath Agent Unit cost in dollars?

UiPath does not publish a dollar value for an Agent Unit or a Platform Unit. The rate is set in your order form and varies with volume, term and negotiation, so any dollar-per-unit figure circulating publicly is a reported range rather than a published price. Size your estimate in units and apply the rate once you have a quote.

Are UiPath agent units poolable across a team?

No. UiPath states that the monthly AI and agentic usage limit is assigned individually to each named user and is not poolable at tenant or organization level. Separately, each tenant is allocated zero agent units by default, and consumption draws from the organization pool until you allocate deliberately.

Does UiPath charge for agent development and testing?

Yes. Agent Units are consumed at design time for debug runs and evaluations as well as at run time in production. This is a common gap in first-year estimates, because during a build quarter development runs can outnumber production runs.

Is bring-your-own-model cheaper on UiPath?

It changes the shape of the cost, which usually matters more than the level. A customer-managed model is charged at 1 Agent Unit per agent run regardless of internal LLM calls, so a twelve-call agent costs the same unit count as a one-call agent. You pay your own inference bill instead. For multi-step agents that is both cheaper and far more predictable, and it requires the Enterprise tier.

Can I set a spending cap on a UiPath agent?

You can cap how often it acts, not how much it spends. UiPath's policy engine caps tool calls and LLM calls per session and per minute, which limits your Agent Unit bill. There is no monetary field anywhere in the platform's public SDKs, so an agent that holds a card or a payment API needs a dollar limit set outside UiPath.