Licensing

Licensing model for Agents, covering agent runs, plan entitlements, and how consumption maps to Platform Units (Unified Pricing) or agent units (Flex).

Understanding agent runs

An agent run refers to a complete execution of an agent, tracked under a single run ID. Each run starts with a trigger and continues until it reaches a terminal state: Succeeded, Failed, Canceled, or Timed-out.

Licensing and consumption

• Production runs consume platform units or agent units, depending on your pricing plan.
• Design-time tests performed in Studio Web consume only LLM quota — no platform units or agent units are consumed.

Note:

When running agents from a shared folder, an unattended robot identity must be assigned to the folder. This is an identity requirement, not a licensing requirement — the agent itself does not consume an unattended robot license during development, debugging, or production runs. For setup details, refer to Assigning objects to a folder.

What counts as an agent run

• Runs invoked from Orchestrator after publishing (manual start, schedule, or queue trigger).
• Runs invoked via API or webhook in a production workspace.
• A conversation session started on a published channel (counts as a single run until it ends or idles out).

What does not count as an agent run

• Studio Web Test/Preview during agent build or design time.
• Autopilot, score runs, and simulations/evaluations marked as preview.

Unified Pricing Licensing

Important:

The information on this page applies to you if you are on Unified Pricing. If you are on Flex, refer to Flex Licensing. For general licensing information, refer to the Automation Cloud Admin guide.

This page provides details on the licenses that give you access to agents and how agent activities are quantified through Platform Units.

Agent licenses and entitlements

The following tables provides a list of licenses that give you access to agents, the plans associated with each license, and the entitlements enabled by them.

Table 1. Licensing plans

Feature	Free	Basic Trial	Basic	Standard Trial	Standard	Enterprise	Application Test Standard Trial	Application Test Standard	Application Test Enterprise
Develop agents: Design, conduct debug runs, and evaluate agents in Studio Web	❌	✅	✅	✅	✅	✅	✅	✅	✅
Run agents: Deploy and run agents as processes	❌	❌	❌	✅	✅	✅	✅	✅	✅

Table 2. User licenses

Feature	Express User	Basic User	Plus User	Pro User	App Tester	App Tester Developer
Develop agents: Design, conduct debug runs, and evaluate agents in Studio Web	❌	✅	✅	✅	❌	✅
Run agents: Deploy and run agents as processes	❌	✅	✅	✅	✅	✅

Design time entitlements

Enterprise

All Enterprise licensed users receive a monthly AI and Agentic usage pool, shared across UiPath AI and Agentic products and usable for agent development and evaluations. The pool replaces the daily entitlement of free agent runs previously granted as LLM calls.

When you use the pool for agent design time, usage is drawn down based on the cost of the model you use and the complexity of the LLM calls executed, counting both input and output tokens.

The monthly limit is available per the following rules:

• The monthly limit is shared across all AI and Agentic products, not only the Agent builder experience in Studio Web. It is not granted in the form of Platform Units.
• The monthly limit is assigned individually to each named user and is not poolable at tenant or organization level.

For details on how the monthly limit, top-ups, and paid-unit fallback work, refer to AI and Agentic usage pool.

Note:

The AI and Agentic usage pool requires User Licensing to be enabled for your organization. When User Licensing is not enabled, all agent interactions at design time are charged per the standard consumption logic.

The pool is available on the Unified Pricing and Flex licensing plans, and is not available on Community.

Feature availability also depends on the cloud platform that you use; for details, refer to the Feature availability page.

The size of the monthly limit depends on the user license, as follows.

User license	AI and Agentic usage pool
Basic User	Basic Usage Pool
Plus User	Basic x7.5
Pro User	Basic x20
App Tester	Basic x7.5
App Tester Developer	Basic x20

The consumption of pool and purchased entitlements during design time and run time occurs according to the following scenarios:

Platform units availability	Design time execution	Run time execution
No purchased units	Allowed up to the monthly limit. Blocked once the monthly limit is reached.	Not allowed
Purchased and available units	Consumes from the monthly limit first. Once the monthly limit is reached, consumes Platform Units. If an administrator grants a top-up, consumption draws from the top-up.	Allowed from available Platform Units

Community

Each Community organization receives a fixed number of LLM calls per organization daily, as follows:

• 250 LLM calls if using UiPath-managed models
• 350 LLM calls if using customer-managed models. For details, see Configuring LLMs in the AI Trust Layer documentation.

Note that community users do not have access to preview features.

Platform Units consumption

Platform Units measure the usage of agent-specific activities. Any agent execution consumes Platform Units, at both design time (for debug and evaluations) and run time.

For UiPath-hosted models, consumption is based on the number of LLM calls made during an agent run, with rates varying by model tier.

For customer-managed models, consumption is charged once per agent run at the Agent Rate, regardless of the number of LLM calls made within that run.

Table 3. Agents consumption rate per model tiers

Model tier	Standard consumption rate	How consumption is calculated
Premium models (preview): Claude Opus 4.7, GPT-5.5	0.4 Platform Units (Agent Rate × 2)	Charged per LLM call
Standard models: GPT-4.1, GPT-4o, GPT-5, GPT-5.1, GPT-5.2, GPT-5.4, Claude 3.5 Sonnet, Claude 3.7 Sonnet, Claude Sonnet 4, Claude Sonnet 4.5, Claude Sonnet 4.6, Gemini 3 Pro, Gemini 3.1 Pro (preview)	0.2 Platform Units (referred to below as Agent Rate)	Charged per LLM call
Basic models: Claude Haiku 4.5, Gemini 2.0 Flash, Gemini 3 Flash (preview), GPT-4.1-mini, GPT-4o-mini, GPT-5-mini	0.16 Platform Units (Agent Rate × 0.8)	Charged per LLM call
Customer-managed models (through the Bring your own model/subscription feature in AI Trust Layer)	0.2 Platform Units (Agent Rate)	Charged once per agent run

Note:

We charge for LLM calls in 64k input token increments. Input tokens refer to the content sent to the model per request — including the system prompt, conversation history, tool definitions, and user messages — and do not include the model's response. For calls exceeding this limit, multiple charges apply. For example, a 100k input token call incurs two LLM call charges. This token limit does not apply to customer-managed models.

You can work with your sales representative to assess your agent use case and predicted Platform Units consumption.

Flex Licensing

Note:

The information on this page applies to you if you are on the Flex Plan. If you are on Unified Pricing, refer to Unified Pricing Licensing. For details, refer to Licensing plan framework.

This page provides details on the licenses that give you access to agents and how agent activities are quantified through agent units.

Agent licenses and entitlements

The following table provides a list of licenses that give you access to agents, the plans associated with each license, and the entitlements enabled by them.

License	License plan	Entitlements
Unlicensed	Community 1	Develop agents: Design, conduct debug runs, and evaluate agents in Studio Web
Automation Developer	Enterprise	Develop agents: Design, conduct debug runs, and evaluate agents in Studio Web Run agents: Deploy and run agents as processes
Citizen Developer
Cloud Basic User

¹ – Community users can develop agents using the design time entitlements.

Design time entitlements

Enterprise

All Enterprise licensed users receive a monthly AI and Agentic usage pool, shared across UiPath AI and Agentic products and usable for agent development and evaluations. The pool replaces the daily entitlement of free agent runs previously granted as LLM calls.

When you use the pool for agent design time, usage is drawn down based on the cost of the model you use and the complexity of the LLM calls executed, counting both input and output tokens.

The monthly limit is available per the following rules:

• The monthly limit is shared across all AI and Agentic products, not only the Agent builder experience in Studio Web. It is not granted in the form of agent units.
• The monthly limit is assigned individually to each named user and is not poolable at tenant or organization level.

For details on how the monthly limit, top-ups, and paid-unit fallback work, refer to AI and Agentic usage pool.

Note:

The AI and Agentic usage pool requires User Licensing to be enabled for your organization. When User Licensing is not enabled, all Agent actions at design time are charged per the standard consumption logic.

The pool is available on the Unified Pricing and Flex licensing plans, and is not available on Community.

Feature availability also depends on the cloud platform that you use; for details, refer to the Feature availability page.

The size of the monthly limit depends on the user license, as follows.

User license	AI and Agentic usage pool
Cloud Basic User	Basic Usage Pool
Citizen Developer	Basic x7.5
Automation Developer	Basic x20

The consumption of pool and purchased entitlements during design time and run time occurs according to the following scenarios:

Agent units availability	Design time execution	Run time execution
No purchased units	Allowed up to the monthly limit. Blocked once the monthly limit is reached.	Not allowed
Purchased and available units	Consumes from the monthly limit first. Once the monthly limit is reached, consumes agent units. If an administrator grants a top-up, consumption draws from the top-up.	Allowed from available agent units

Community

Each Community organization receives a fixed number of LLM calls per organization daily, as follows:

• 250 LLM calls if using UiPath-managed models
• 350 LLM calls if using customer-managed models. For details, see Configuring LLMs in the AI Trust Layer documentation.

Note that community users do not have access to preview features.

Agent unit consumption

Agent units measure the usage of agents-specific activities. Any agent execution consumes agent units, at both design time (for debug and evaluations) and run time.

Agent unit consumption depends on the model type used by the agent and differs between UiPath-hosted and customer-managed models.

By default, each tenant is allocated 0 agent units, and agent unit consumption draws from the organization pool. If agent units are allocated to a specific tenant, that tenant's allocation is used for consumption.

Consumption is calculated as follows:

• UiPath-hosted models: Consumption is based on the number of LLM calls made during an agent run. Each LLM call is charged as an agent action, and multiple calls within a single run result in multiple charges.
• Customer-managed models (Bring Your Own Model or Bring Your Own Subscription): Consumption is charged once per agent run, regardless of the number of LLM calls made within that run. LLM calls made using customer-managed models are not metered.

Table 4. Agents consumption rate per model tiers

Model tier	Standard consumption rate	How consumption is calculated
Premium models (preview): Claude Opus 4.7, GPT-5.5	2 Agent Units	Charged per LLM call
Standard models: GPT-4o, GPT-4.1, GPT-5, GPT-5.1, GPT-5.2, GPT-5.4, Claude 3.5 Sonnet, Claude 3.7 Sonnet, Claude Sonnet 4, Claude Sonnet 4.5, Claude Sonnet 4.6, Gemini 3 Pro, Gemini 3.1 Pro (preview)	1 Agent Unit	Charged per LLM call
Basic models: Claude Haiku 4.5, Gemini 2.0 Flash, Gemini 3 Flash (preview), GPT-4.1-mini, GPT-4o-mini, GPT-5-mini	0.8 Agent Units	Charged per LLM call
Customer-managed models (through the Bring your own model/subscription feature in AI Trust Layer)	1 Agent Unit	Charged once per agent run

Note:

For UiPath-hosted models, we charge for LLM calls in 64k input token increments. Input tokens refer to the content sent to the model per request — including the system prompt, conversation history, tool definitions, and user messages — and do not include the model's response. For calls exceeding this limit, multiple charges apply. For example, a 100k input token call incurs two LLM call charges. This token limit does not apply to customer-managed models. For customer-managed models (BYOM/BYOS), LLM calls are not metered, as consumption is charged once per agent run.

Agent units consumption is displayed in Automation Cloud Admin, on the Consumables tab of the tenant-level Licensing page. For details, refer to The Consumables tab.

You can work with your sales representative to assess your agent use case and predicted agent unit consumption.