UiPath AI Agents Pricing: What It Really Costs You (2026)

---

title: "UiPath AI Agents Pricing: What It Really Costs You (2026)" slug: "uipath-ai-agents-pricing-what-does-it-cost-you-2025" date: "2026-08-05" excerpt: "UiPath's published agent metering rates, the Flex vs Unified Pricing framework, worked cost math, and honest 2026 comparisons against Automation Anywhere, Microsoft, and IBM." author: "O-mega Team" tags:

• UiPath
• AI Agents
• Pricing
• Agentic Automation
• Platform Units
• RPA category: "Automation Technology" seoTitle: "UiPath AI Agents Pricing: What It Really Costs You (2026)" seoDescription: "UiPath agent pricing in 2026: exact Platform Unit rates per LLM call, Flex vs Unified Pricing, worked cost scenarios, and verified competitor comparisons."

---

The practitioner's breakdown of what UiPath's AI agents actually cost in 2026, with the published metering rates no aggregator page does the math on.

Two pricing guides currently quote "$420" for a UiPath license. One says per year. The other says per month. Automation Atlas estimates an attended robot at roughly $420 per user per year, while PricingNow lists the Automation Developer license at $420 per user per month (about $5,040 annually). That is a 12x disagreement on nearly the same number, published by sites that both claim to explain UiPath pricing. The number itself has a history: UiPath used to sell a fixed Pro plan at $420/month, a plan that no longer exists. The old price point keeps getting recycled into new meanings by aggregators that never checked the source.

That disagreement is the single most useful fact about UiPath pricing, because it tells you what kind of market you are buying in: one where the vendor publishes exact consumption meters but not what the meter charges, and where third parties fill the vacuum with conflicting estimates. This guide takes the opposite approach. Everything quantitative below comes from a source we opened while writing, most of it from UiPath's own licensing documentation, which now publishes the exact per-LLM-call rates for agents. We run the unit-economics math on those rates step by step, something none of the pages currently ranking for this query does.

One disclosure before we start: we build O-mega, a platform where AI builds and runs autonomous companies, so we compete for some of the same budgets UiPath does. We also pay LLM bills ourselves and meter usage in credits, which means we know this pricing model from the vendor side. We will tell you where UiPath is genuinely the right buy, where it is not, and where we are not either.

Contents

1. What changed since 2025 (the deaths and renames)
2. UiPath's 2026 plan structure
3. The agentic product suite, by name
4. How agent metering actually works: Flex vs Unified Pricing
5. The worked scenario: what an email-triage agent consumes
6. Why third-party price estimates disagree, and how to negotiate
7. Honest alternatives: Automation Anywhere, Microsoft, IBM, O-mega
8. Where O-mega fits (first-person, bias declared)
9. Verdict: when UiPath is the right buy, and when it is not

1. What Changed Since 2025 (The Deaths and Renames)

This article first ran in July 2025, and enough of it has since stopped being true that a quiet touch-up would be dishonest. Pricing content ages badly in this market, and the changes themselves are the most valuable information: knowing which products died and which numbers expired saves you from negotiating against a ghost. So before the current structure, here is what expired.

The fixed Pro plan at $420/month, which the 2025 version of this article used as the anchor estimate for a professional UiPath package, is gone: Pro survives only as a per-user license type, not a bundled plan - Automation Atlas. The widely repeated "$4,000 per year per robot" figure, which we also cited, traces back to older aggregator research and no longer matches what 2026 third parties estimate - Monetizely. The vague "AI Units" framing we described as UiPath's current agent currency is now legacy terminology from the Flex plan; UiPath's current framework splits into Flex (separate unit types) and Unified Pricing (one Platform Unit), covered in section 4 - UiPath docs. And the "Unified Pricing initiative aims to simplify" line from 2025 is out of date in the good direction: it shipped, with published per-call rates.

The competitive landscape rotated too. SAP Intelligent RPA, which the old article suggested SAP-centric readers evaluate, was already a dead product name at the time: SAP folded it into SAP Build Process Automation, with a migration path SAP documented publicly - SAP Community. We were wrong at publish, and naming that plainly matters more than the correction itself: it shows how fast "current" vendor lists rot. IBM watsonx Orchestrate moved the other direction, from quote-only to published tiered editions you can buy self-serve - IBM. Microsoft restructured Copilot into seven pricing tiers in July 2026, up from four - CostBench. Each of these gets full treatment in section 7 with only the numbers we verified this run.

2. UiPath's 2026 Plan Structure

The top-level plan structure is the one part of UiPath pricing that has stayed stable, and it is still the first thing a buyer needs to understand because everything else (agent metering, robot licenses, support tiers) hangs off which plan you are on. UiPath's pricing page currently lists Community (Free), Basic, Standard, and Enterprise for Automation Cloud. Only one of those four has a published price.

Basic starts at $25 per month, and the constraints around that number matter more than the number: it covers personal automations at limited scale, is hosted in the European region, and comes with 99.9% uptime and Bronze support - UiPath pricing. Crucially for this article's topic, Basic includes only limited agentic capabilities. Standard is quote-only and is where "enterprise automations with agents" begins: full agentic automation support, document classification and extraction, process orchestration, enhanced governance, and unlimited scale. Enterprise, also quote-only, layers on self-healing UI automation, custom AI models, encryption key management, and credential vault management. The pattern to internalize: the $25 entry price is real but is not the product this article is about. Agents live in the quote-only tiers.

There is still a genuinely useful free on-ramp. The Automation Cloud trial runs 60 days and can be extended by another 30 or 60 days on request, and it includes 5 attended, 5 unattended, and 5 test robots plus Studio, Document Understanding, AI Center, and Action Center - UiPath docs. For a team seriously evaluating agentic automation, 60 to 120 days with that allocation is enough to build and measure a real pilot rather than a demo. Community organizations also get a daily pool of free agent LLM calls (exact numbers in section 4), which means you can prototype an agent's consumption profile before you ever talk to sales, and we strongly recommend doing exactly that: walk into the pricing conversation with your own measured call counts.

So when do you actually need to start paying? The threshold question the 2025 version of this article answered with generalities now has a numeric answer. The Community pool of 250 free LLM calls per day on UiPath-managed models supports roughly 50 agent runs a day at the 5-calls-per-run shape we model in section 5, which genuinely covers a proof of concept but not a production inbox - UiPath docs. The moment your agent needs to run unattended in production, against real business data, at volumes above that free pool, you are in quote territory: the Standard tier is where "enterprise automations with agents" formally begins - UiPath pricing. The practical sequencing that follows from the license mechanics: exhaust the trial and free allowances while measuring consumption, and only then open the sales conversation, because the free tiers are the only phase in which UiPath's meter runs without a negotiated price attached, and the data they generate is your leverage in that negotiation.

3. The Agentic Product Suite, By Name

The 2025 version of this article discussed "UiPath AI Agents" as an abstract capability, which was already behind reality then and is useless as buyer vocabulary now. When you get a UiPath quote in 2026, it will name specific products, and you should know what each one is before the line items arrive. This section exists because none of the pricing pages currently ranking for this query names them at all.

The platform launched formally on April 30, 2025, when UiPath announced what it called the first enterprise-grade platform for agentic automation, unifying AI agents, robots, and people in one system - UiPath newsroom. The two names that matter most for cost conversations are Agent Builder and Maestro. Agent Builder is where agents get made: it lets teams prototype and deploy agents inside UiPath Studio, with low-code assembly for simple cases and pro-code customization for complex ones like invoice dispute resolution - UiPath. Maestro is the orchestration layer: UiPath describes it as automating, modeling, and optimizing complex business processes end to end, with built-in process intelligence and KPI monitoring, coordinating agents, robots, and humans across long-running processes. The launch also included IXP (Intelligent Xtraction and Processing) for multi-modal document work and a UI Agent for computer use, then in private preview.

Why this matters for pricing: these products are what the quote-only tiers are selling, and they consume the metered units differently. An Agent Builder agent burns LLM calls (metered per call, next section), while Maestro orchestration is part of the platform entitlement you negotiate at the Standard or Enterprise level. When a UiPath seller talks about "agentic transformation," the concrete question to ask is: which products, consuming which units, at what rate. The rest of this guide gives you the rates. For background on how orchestration layers like Maestro relate to the broader agent stack, our guide to workflow automation with AI agents covers the architecture patterns these products implement.

4. How Agent Metering Actually Works: Flex vs Unified Pricing

This is the section that did not exist in 2025 because UiPath had not yet published the numbers. It has now, and the exact rates are the core asset of this article. UiPath prices agent execution by metering LLM calls, and the rate per call depends on two things: which commercial framework your contract uses, and which model tier the agent calls. Both are documented publicly, which is more transparency than most agent vendors offer, including, historically, us.

First, the framework split. UiPath contracts now come in two shapes - UiPath docs. The Flex plan uses separate purpose-specific consumables: AI Units, Apps Units, Robot Units, API Calls, and Agent Units, each bought and burned independently. Unified Pricing replaces all of those with a single Platform Unit consumable that any UiPath service can draw from, and, importantly, allows reallocation: if you bought units expecting document processing and ended up running agents instead, the units move with your plans. That reallocation right is the practical reason to prefer Unified Pricing in negotiation: with Flex, a mis-forecast strands budget in the wrong unit type.

Second, the rates. Under Unified Pricing, agent execution consumes Platform Units per LLM call at rates set by model tier - UiPath licensing docs. The premium tier, which at the time of writing UiPath's documentation populates with Claude Opus 4.7 and GPT-5.5, costs 0.4 Platform Units per call. The standard tier costs 0.2 Platform Units per call. The basic tier, holding the small models, costs 0.16 Platform Units per call. Customer-managed models (you bring your own model endpoint) are charged at 0.2 Platform Units once per agent run rather than per call, which quietly makes bring-your-own-model the most predictable line on the whole rate card. Flex contracts meter the same tiers in Agent Units at exactly five times the Platform Unit figures: 2 per premium call, 1 per standard call, 0.8 per basic call. The 5:1 ratio is constant across every tier, which tells you the two frameworks are the same meter wearing different denominations.

Two fine-print rules change the math more than the headline rates. First, calls are metered in 64k input-token increments for UiPath-hosted models: a single logical call whose input exceeds 64k tokens is charged as multiple calls - UiPath docs. Feed an agent long documents, big retrieval contexts, or sprawling email threads and your effective per-call cost multiplies without the call count changing. Second, the model tier assignments are UiPath's, not the market's. The documentation's standard tier currently spans models across several generations, from GPT-4o (a 2024-era model) through recent Claude Sonnet and Gemini Pro releases, all billed at the same 0.2 rate, while only two models sit in premium. Tier placement will shift as models rotate, so re-check the doc at contract time rather than trusting any article, including this one.

There is also a meaningful free allowance, and it is worth planning around. On Enterprise contracts, each user license carries free daily agent LLM calls at design time: a Pro or Automation Developer license gets 2,500 calls per day (roughly 500 agent runs), a Plus or Citizen Developer license gets 125 calls (about 25 runs), and a Basic or Cloud Basic user gets 50 calls (about 10 runs) - UiPath docs. Community organizations get a daily pooled allowance of 250 LLM calls on UiPath-managed models, or 350 if the models are customer-managed. The design-time qualifier is the catch: these allowances cover building and testing agents, while production runs draw from your paid units.

Now the honest limitation, which is also the strategic core of UiPath's pricing: the dollar price of a Platform Unit is not published. UiPath publishes the meter with admirable precision and keeps the meter's price inside the quote. You can compute your workload's unit consumption to the decimal (next section does exactly that) but you cannot convert units to dollars until sales gives you your number, and your number depends on volume, term, and negotiation. Why do vendors meter this way? Because agent workloads have wildly variable underlying costs: per-call metering passes model-choice and prompt-length risk to the buyer, while the unpublished unit price preserves price discrimination between customers. We meter our own product by credits for the first reason, so we say this without pretending to be above it: every agent vendor's meter is a cost-passthrough plus a margin you cannot see. The difference between vendors is only how much of the meter they publish. UiPath publishes the rates; the unit price is the negotiation.

5. The Worked Scenario: What an Email-Triage Agent Consumes

Rates in the abstract do not answer the question a buyer actually has, which is "what will my use case consume?" So let us run one, end to end, using only UiPath's published numbers. This is math you can redo for your own workload in ten minutes, and doing it before the sales call is the single highest-leverage preparation available to you, because it converts the conversation from "trust our sizing" to "price this consumption."

Take a realistic first agent: email triage for a shared inbox. The agent reads each incoming message, classifies it, drafts a response or routes it with context. Assume 500 emails per day, and assume each run makes about 5 LLM calls (classify, retrieve context, reason, draft, verify): a typical shape for a tool-using agent. On a standard-tier model at 0.2 Platform Units per call, that is 500 runs x 5 calls x 0.2 units = 500 Platform Units per day - rates. Run it every working day and you are near 11,000 Platform Units per month on one modest agent. Switch the same agent to a premium-tier model at 0.4 and daily consumption doubles to 1,000 units; drop to the basic tier at 0.16 and it falls to 400.

Now apply the fine print, because this is where real bills diverge from clean scenarios. The 64k input-token increment rule means the "5 calls" assumption only holds while inputs stay short. An email thread with a long history, a PDF attachment run through extraction, or a fat retrieval context can push individual calls over the increment, and each overflow bills as an additional call. In our own operations at O-mega, long-context steps are routinely the difference between an agent costing what we modeled and costing multiples of it; prompt length is a cost driver buyers systematically underestimate because no dashboard shows it before launch. The general version of this problem, and the techniques for containing it, are covered in our guides to the true cost of LLM inference and cutting LLM costs, both of which apply directly to UiPath agent workloads even though they were not written about UiPath.

Three design levers fall straight out of the arithmetic, and they are worth naming because they are contract-relevant, not just engineering-relevant. Model tier selection is a 2.5x swing (0.16 to 0.4) at identical call counts, so insisting that routine steps run on basic-tier models is the cheapest optimization available; the same logic drives model routing, where cheap models handle easy steps and expensive ones handle judgment. Calls per run is your prompt-engineering budget: collapsing five steps into three cuts consumption 40% with no quality change if the collapsed steps were genuinely mechanical. And customer-managed models, charged once per run rather than per call, decouple UiPath's meter from your model bill entirely: you pay your own inference provider for tokens and UiPath a flat 0.2 units per run for the privilege of orchestration. For high-call-count agents, that structure can dominate both alternatives, and UiPath's own rate card is what makes it visible.

What the scenario cannot tell you is dollars, and we will not fake it: no published Platform Unit dollar price exists to multiply against that 11,000-unit month. What it gives you instead is the negotiation posture. You walk in saying "our first agent consumes roughly 11,000 Platform Units a month on standard-tier models; quote me the unit price at that volume, and at 10x that volume," and the quote you get back is now comparable, auditable, and anchored to your math instead of their sizing exercise.

6. Why Third-Party Price Estimates Disagree, and How to Negotiate

If you searched "UiPath pricing" before landing here, you met a wall of confident, conflicting numbers. It is worth understanding why they conflict, because the mechanism tells you which numbers deserve any weight at all. Quote-based enterprise pricing means every published estimate is either a leaked contract (one data point, unknown discount), a reseller's rule of thumb (regional, dated), or an aggregator recycling another aggregator. Errors then compound: a real number from one year becomes a different license's price the next, as the $420 story from our introduction shows, where one 2026 guide has it as an attended user's annual cost and another as a developer's monthly cost - Automation Atlas vs PricingNow.

With that lens, here is the current spread, clearly attributed and not averaged, because averaging conflicting estimates manufactures false precision. Automation Atlas estimates ~$1,680 per unattended robot per year, ~$420 per attended user per year, Orchestrator at $8,000 to $20,000 per year, and AI Center at $5,000 to $15,000, and illustrates with a 300-person healthcare deployment totaling $19,140 per year - Automation Atlas. PricingNow models the Automation Developer license at $420 per user per month, scaling to $50,400 a year for ten users - PricingNow. At program level, Forrester's range for enterprise automation platforms in this class is $100,000 to several million dollars annually, as cited by Monetizely. Those figures are not reconcilable into one price card, and that is the point: treat every third-party UiPath number, including the ones in this paragraph, as a negotiation reference, not a price.

If you need budget bands rather than point estimates, the honest ones available in 2026 are deliberately wide. Nerova's UiPath cost guide declines to publish per-robot figures at all and instead frames a realistic business pilot at "a few hundred to a few thousand dollars per month," a department rollout "in the low thousands to low five figures per month," and an enterprise program as "a six-figure-plus annual program" - Nerova. The same guide makes the structural observation that matches our own operating experience: a program leaning heavily into agentic workflows, AI extraction, or hosted robot capacity has monthly spend that is "less predictable than a pure seat-based model." That unpredictability is not a UiPath defect; it is the nature of consumption-priced AI, and it is exactly why the measured-consumption negotiation posture below matters more here than in classic seat-license deals.

The negotiation itself rewards preparation more than most software deals, precisely because so little is published. The levers that consistently matter are worth spelling out in prose rather than a checklist, because each has a reason. Measured consumption beats estimated consumption: pilot on the 60-day trial and Community free calls, arrive with your own unit math from section 5, and make the unit price the negotiation instead of the bundle. Framework choice is negotiable: Unified Pricing's reallocatable Platform Units protect you from forecast error in a way Flex's earmarked units do not, so price both. Model-tier governance belongs in the contract conversation: an agent fleet defaulting to premium-tier models doubles consumption against standard, so ask what controls exist to pin workloads to cheaper tiers. Term-versus-flexibility is the classic trade: multi-year commitments buy better unit prices exactly when your consumption forecast is least reliable, which is early in an agent program; shorter terms cost more per unit and are often worth it. And the platform-tax question should be asked in exactly these words: what do we pay before the first agent runs in production, counting Standard-tier entry, user licenses, and any Orchestrator or infrastructure components.

One more honest note about sources, since this article is itself a source: earlier versions of this piece cited research.aimultiple.com figures like the $4,000-per-robot-per-year estimate that still circulates - Monetizely. Those aimultiple research URLs now redirect to a reorganized site, and the numbers predate the agentic repricing entirely. If a page quotes UiPath robot prices without a 2026 source, you are reading the fossil record. Our own 2024 UiPath pricing breakdown is preserved as exactly that: a snapshot of the pre-agentic license model, useful now mainly to measure how much the model changed.

7. Honest Alternatives: Automation Anywhere, Microsoft, IBM, O-mega

Comparison sections in pricing articles are usually where the padding lives, so this one follows a strict rule: one row per vendor, only verified numbers, every cell sourced, and where a vendor publishes nothing, the table says so instead of inventing a figure. Entry price means the cheapest published way to start paying; it deliberately does not mean "what you will pay," which for every enterprise vendor here is a negotiated number.

The vendors are not interchangeable, which is why this is a reference table rather than a scored ranking. UiPath and Automation Anywhere are enterprise automation platforms with agents added to an RPA core. Microsoft sells per-user and per-bot licensing attached to its ecosystem. IBM sells orchestration of agent catalogs. We sell autonomous company-building. Ranking those on one scale would manufacture comparability that does not exist; what a buyer actually needs is each vendor's pricing shape, verified.

Automation Anywhere remains UiPath's most direct competitor: same enterprise-RPA heritage, same agent pivot, and its agent tooling is AI Agent Studio, the low-code agent design layer inside Automation 360. Its pricing went the opposite direction from UiPath's meter-publishing: fully quote-based, with only the free Community edition public. The best available price signal is third-party benchmark data, and it should be labeled as exactly that: CostBench reports a median annual contract of $27,650 across 10 verified purchases, with Vendr-derived typical spend between roughly $11,000 and $72,000 and large deployments exceeding $750,000 per year - CostBench. That is benchmark data from real buyers, not vendor pricing. The $750/month bundle figure this article quoted in 2025 has no current source and should be considered dead. Our older Automation Anywhere pricing breakdown documents the earlier license-era structure for comparison.

Microsoft is the transparency outlier, publishing nearly every relevant number, which is why it anchors so many automation budget conversations. Power Automate Premium is $15 per user per month; the Process plan (unattended RPA) is $150 per bot per month; the newer Hosted Process plan, where Microsoft runs the bot infrastructure, is $215 per bot per month, all billed annually - Zapier. On the Copilot side, Microsoft 365 Copilot remains $30 per user per month paid yearly as the enterprise add-on - Microsoft. The July 2026 restructuring expanded the lineup from four tiers to seven - CostBench. The small-business SKUs now run Copilot Business at $18 per user per month annual (a promotional rate against a $21 list), Business Standard with Copilot at $23.50, and Business Premium with Copilot at $32 - Microsoft. Worth knowing before you anchor on list prices: CostBench's verified-purchase median for Copilot contracts is $417 per year, well under list, meaning even Microsoft's published prices get negotiated in practice. Our Power Automate pricing guide tracks the per-user model's earlier iterations.

IBM watsonx Orchestrate is the vendor the 2025 version of this article got most wrong going forward: we wrote "exact figures aren't public," and IBM has since productized pricing into tiered editions: a free trial, Essentials for early teams, and Standard for scaling with pre-built agents, purchasable self-serve through the IBM Cloud catalog and AWS Marketplace rather than exclusively through sales - IBM. The platform's 2026 shape is notably agent-forward: agent builder tooling, imports from frameworks like Langflow and LangGraph, an agent gateway with MCP server support, and even an eCommerce pilot selling partner-built agents directly. IBM's page presents the editions without dollar figures (rates surface in the cloud catalogs at purchase), so we will not print a number we could not open a source for; the structural fact is the news: IBM moved from quote-only to published self-serve tiers, which pressures every quote-only vendor in this table.

SAP gets a paragraph so nobody repeats our mistake: if your automation shortlist still says "SAP Intelligent RPA," that product is gone as a standalone offering, consolidated into SAP Build Process Automation, which carries the RPA capability inside SAP's Build portfolio - SAP Community. For deep SAP estates it remains the path of least resistance precisely because it lives inside the SAP stack, and that is also its boundary: it is an SAP-world tool, not a general agentic platform.

8. Where O-mega Fits (First-Person, Bias Declared)

We are in the comparison table, so this section is written in first person and you should apply the same discount to it that you would apply to any vendor writing about itself. O-mega is not an RPA platform and does not try to be one. Our platform builds and runs autonomous companies: you describe what you want in conversation, and the AI builds and operates the company end to end, from the public website through the customer-facing product, billing, email campaigns, and the internal admin dashboard, deployed and live - O-mega plans. The 2025 version of this article described us as a "productivity platform for multi-agent teams" with per-action credits; that description is stale, ours, and now corrected, which is its own small proof of how fast self-descriptions rot in this market.

Our pricing model is the part most relevant to this article's theme, because it is the same consumption logic as UiPath's with a different unit of account. One credit equals one step toward building your company: adding a banner, changing styling, setting up automated emails, publishing a blog post - plans. When credits run out, building pauses by default; overage is strictly opt-in, billed at per-credit rates that improve with plan tier. Enterprise plans start from $25,000 per year with custom setup, dedicated support, and enhanced security. We meter in steps rather than LLM calls because we would rather absorb prompt-length variance ourselves than pass it to the customer invisibly; the trade-off, stated honestly, is that a "step" is our judgment call the way a "call" is UiPath's, and any consumption meter ultimately asks you to trust the meter. Our full analysis of agent cost structures across the industry, including our own, is in the true cost of agentic AI.

When should you not pick us? If your problem is automating existing legacy systems (SAP screens, Citrix sessions, decades-old ERPs, regulated on-prem workflows) UiPath is genuinely better at that than we are; that is what two decades of RPA tooling buys, and we do not pretend a greenfield company-builder replaces it. If your problem is deploying assistants inside Microsoft 365, buy Copilot. We are the right call when the goal is to stand up something new that runs itself, rather than to wrap automation around what already exists.

9. Verdict: When UiPath Is the Right Buy, and When It Is Not

Strip away the unit systems and the verdict question is structural: UiPath's pricing model rewards buyers who already have mass on the platform and punishes buyers who bring none. Every component you already own (robot licenses, Orchestrator, trained developers, governance sign-offs) makes the agentic layer incremental; every component you lack is platform tax paid before the first agent triages its first email. So the honest recommendation splits cleanly by where you start, not by how good the technology is.

UiPath is the right buy when you have an existing UiPath RPA estate whose licenses and skills the agent layer extends; when compliance or on-premises requirements demand the controls its Enterprise tier is built around (custom AI models, encryption key management, credential vaults) - UiPath pricing; and when you genuinely need Maestro-scale orchestration of agents, robots, and humans across long-running processes rather than a handful of standalone agents. In that situation, its published metering is a strength: run the section 5 math, negotiate the unit price on Unified Pricing terms, and you will know your costs better than buyers of most competing platforms can.

UiPath is the wrong buy for greenfield, agent-first teams with no RPA estate, because the quote-only Standard entry plus user licensing plus infrastructure is a tax paid before any agent runs, while Microsoft starts at a published $15 per user and platforms like ours start free. It is also the wrong buy if you cannot staff the platform: the third-party TCO estimates in section 6 diverge, but none of them is small relative to a team of three. And if your automation lives entirely inside SAP or Microsoft ecosystems, the native options (SAP Build Process Automation, Power Automate with Copilot) will be cheaper and closer to your data. The broader pattern, RPA platforms repricing themselves around agents while agent-native platforms attack from below, is the subject of our RPA practical guide, and UiPath's 2026 rate card is that pattern's clearest artifact: the meter is public, the price is not, and the buyer who does the unit math first wins the negotiation.

For deciding this week: pilot on the 60-day trial with your real workload, measure calls per run, apply the published rates, and get quotes for both commercial frameworks at your measured volume and at ten times it. If the resulting number beats what your measured workload would cost on a per-user Microsoft stack or an agent-native platform, buy it; if the platform tax exceeds the first year's automation value, you have your answer just as clearly.

Written by the team at O-mega, with the pricing research led by Yuma Heymans (@yumahey), founder of O-mega and co-founder of HeroHunt.ai, who reads LLM metering bills from the vendor side weekly and wrote the unit math in section 5 the same way he models it for O-mega's own credit system.

This guide reflects UiPath and competitor pricing as of August 5, 2026. Every rate and price above was verified against a live source on that date; consumption rates, model tiers, and SKUs in this market change quarterly, so re-verify against the linked primary sources before signing anything.