- 入门指南
- UiPath Conversational Agents
- Studio Web 中的 UiPath 智能体
- UiPath 编码智能体
- 使用编码智能体构建
Hybrid licensing model for chat agent executions, covering consumption factors, available plans, and deployment planning scenarios.
This page covers licensing for chat agent executions, which is either:
- A single turn of a conversational agent built in Agent Builder
- A single turn (node run) of a conversational agent node within Maestro Flow
Licensing uses a hybrid user- and consumption-based model. This page explains how licensing works, what factors affect consumption, and provides scenarios to help you plan your deployment.
For the Maestro Flow build path, the execution consumption described here is in addition to licensing for the Maestro Flow that hosts the conversational agent node(s).
定价模型概述
UiPath offers two licensing plans for chat agents. Both use a consumption-based model where usage is measured in executions.
What counts as an execution
An execution is a single conversational agent turn, and includes:
- 智能体的响应。
- 两者之间执行的任何工具调用。
Exception: executions do not draw from the monthly limit when the most recent user's message contains eight characters or fewer and no tool calls are executed.
AI 和智能体使用情况池
All Enterprise licensed users receive a monthly AI and Agentic usage pool, shared across UiPath AI and Agentic products and usable for conversational agents. The pool replaces the previous per-user entitlement of 50 free executions per month.
When you use the AI and Agentic usage pool for conversational agents, usage is drawn from the pool per execution, based on the cost of that execution. The pool capacity varies by user license, as shown in the following tables.
有关每月限额、充值以及按单位付费回退工作原理的详细信息,请参阅“AI 和代理使用池” 。
AI 和代理使用量池在 Unified Pricing 和 Flex 许可计划中可用,在 Community 版上不可用。功能可用性还取决于您使用的云平台;有关详细信息,请参阅功能可用性页面。
Flex 许可
Flex licensing uses Agent Units for consumption. When the monthly limit and any top-up are spent, conversational agents continue by drawing from Agent Units, at 1 Agent Unit per execution.
| 用户许可证 | AI 和智能体使用情况池 |
|---|---|
| Express 用户(原名 Automation Express 和 Autopilot Express) | No monthly limit; 50 free executions per user per year, then 1 Agent Unit per execution |
| Cloud Basic User | 基本使用量池 |
| Attended - Named User | 基础 x7.5 |
| Citizen Developer | 基础 x7.5 |
| Automation Developer | 基础 x20 |
统一定价
Unified Pricing 使用Platform Units计算消耗量。当每月限额及所有充值用尽后,对话智能体将继续以对话智能体消耗率提取 Platform Units。
| 用户许可证 | AI 和智能体使用情况池 |
|---|---|
| Basic 用户 | 基本使用量池 |
| Plus 用户 | 基础 x7.5 |
| Pro 用户 | 基础 x20 |
| 应用程序测试者 | 基础 x7.5 |
| 应用测试者开发者 | 基础 x20 |
回退提取的速率因为智能体配置的 LLM 模型层级而异:
| 模型层 | 消耗率 |
|---|---|
| 标准模型 | 0.2 Platform Units(也被称为“对话智能体费率”) |
| 基础模型 | 0.16 Platform Units(对话智能体费率 × 0.8) |
Autopilot for Everyone
Autopilot for Everyone 还可以从每月的 AI 和代理型使用量池中提取内容。在用完每月限额并用完充值后,将按每次操作的对话智能体费率(在 Unified Pricing 时,每次操作消耗 0.2 个 Platform Units)。有关详细信息,请参阅Autopilot for Everyone 许可。
外部用户(无许可证)
对于匿名访问身份验证:
- 不需要用户许可证。
- 需要外部应用程序令牌(在管理门户中创建)。
- Each execution consumes from the tenant's unit pool starting with the first execution.
- 通过 iFrame 嵌入和身份验证服务器进行部署。
设计时权限
Development and debugging of agents draws from the same monthly AI and Agentic usage pool as published agent executions. Design time usage is no longer measured as a separate daily LLM call allotment, because the monthly limit is shared across all AI and Agentic products and activities.
Execution consumption by user license
下表显示了消耗量在月度上限内和上限以外的计算方式:
| 用户许可证 | 在每月限额内 | 达到每月限额后 |
|---|---|---|
| Express 用户 | 50 free executions per user per year | 从租户池中消耗 |
| 所有其他许可用户 | 在每月限额内提取,并根据许可证调整大小 | 执行任何管理员授予的充值操作,然后使用租户池 |
| 外部用户(无许可证) | 不适用 | Consumes from the tenant pool from the first execution |
工具消耗量
While some tool costs are included in the base execution, certain tools have additional consumption requirements. Tools are billed per activity executed, not per connector configured.
| 工具 | Flex(智能体/其他单元) | Unified (Platform Unit) |
|---|---|---|
| 分析文件 | 每个文件 1 个 AU | 每个文件 0.2 个 PU |
| IXP 工具 | 每页 1 个 AU | 每页 0.2 个 PU |
| MCP 工具 | 包括 | 包括 |
| 上下文锚定(语义搜索) | 包括 | 包括 |
| 上下文锚定 (DeepRAG) | 每 ~20 页 1 个 AU | 每 ~20 页 0.2 个 PU |
| Integration Service 活动 | 包括 | 包括 |
| API 工作流 | 每执行一个活动调用 5 次 IS API | 每执行一个活动消耗 0.02 个 PU |
| 跨平台 RPA | 1 个 Robot Unit | 0.2 个 PU |
| Windows RPA (Unattended) | 需要 Unattended Robot | 需要 Unattended Robot |
| 自主智能体(嵌套) | 嵌套智能体的消耗量 | 嵌套智能体的消耗量 |
场景
以下场景说明了许可在实践中如何运行。
场景 1:简单的问答助手
智能体配置
- 工具:仅上下文锚定语义搜索
- 无需其他工具
消耗量计算
Because this example uses only Context Grounding semantic search, there is no additional tool cost per execution.
Monthly limits are not pooled across users. Each user has their own monthly limit, consumable only by them. If User A uses a fifth of their limit while User B exhausts theirs, User A still has four fifths remaining, and User B's further executions draw from a top-up if an administrator granted one, and then from the tenant pool.
示例:不同许可证类型的使用量:
- User A (Basic User): around 20 executions this month, so roughly 40% of the monthly limit used
- User B (Basic User): around 100 executions this month, so the monthly limit is exhausted after roughly 50 and the remainder draws from the tenant pool
- User C (Pro User): around 500 executions this month, so roughly half of the monthly limit used
- Total tenant pool consumption: only User B's executions beyond their monthly limit
对于希望采用纯消耗式模型且无需维护 Automation Cloud 用户帐户的客户,应通过 iFrame 嵌入,使用匿名访问权限部署对话智能体。这会立即开始消耗单元,无免费配额。
场景 2:有多个工具的自助服务助手
智能体配置
- 上下文锚定语义搜索
- “分析文件”工具
- Integration Service 连接器
- 已执行两个连接器活动的 API 工作流
消耗量计算
其他工具成本
| 交互类型 | 消耗量 (Flex) | 消耗量 (Unified) |
|---|---|---|
| 语义搜索 | 包括 | 包括 |
| 文件分析 | 每个文件 1 个 AU | 每个文件 0.2 个 PU |
| IS 连接器 | 包括 | 包括 |
| API 工作流(已执行 2 个活动) | 10 次 IS API 调用 | 0.04 个 PU |
示例:不同许可证类型和工具的月度使用量:
| 用户 | 许可证 | 执行 | Flex 消耗量 | Unified 消耗量 |
|---|---|---|---|---|
| A | 快速 | 30(问答消耗 20 个,文件分析消耗 10 个) | 30 free executions, 10 AU for file analysis | 30 free executions, 2 PU for file analysis |
| B | 基本 | 75(问答消耗 50 个,文件分析消耗 15 个,API 工作流消耗 10 个) | Executions draw from the monthly limit, then 1 AU each; 15 AU tools + 100 IS API calls | Executions draw from the monthly limit, then 0.2 PU each; 3.4 PU tools |
| C | Plus | 200(混合工具) | Executions draw from the monthly limit, tool consumption applies | Executions draw from the monthly limit, tool consumption applies |
自带模型(预览版)
Conversational agents support LLM Configurations so you can use your own LLM subscriptions. This feature is in preview. During the preview period, there is no charge per execution, however additional consumption based on tool usage still applies.
管理消耗量
管理员控件
管理员可以在租户级别管理消耗量:
- 通过管理门户监控使用量。
- 为消耗阈值设置配额。
- 审核各个智能体的使用量。