Liste de contrôle de la configuration des agents

Follow the steps below to make sure Agents work as expected in Automation Suite:

1. Make sure you meet all prerequisites before you start. For details, refer to Prerequisites.
2. In Automation Ops, create or edit an AI Trust Layer policy and enable the Agents toggle on the Product Toggles tab, then deploy the policy to your tenant. For details, refer to Settings for AI Trust Layer policies.
3. Configure at least one LLM connection in the AI Trust Layer. Agents requires an LLM to function — all AI-powered features are unavailable until an LLM is configured. For details, refer to Configuring LLMs for your product.
4. Build an agent in Agent Builder. For details, refer to Getting started with UiPath Agents.
5. Test the agent to verify it runs as expected. For details, refer to Testing the agent.
6. Publish and deploy the agent to make it available for use. For details, refer to Publishing and deploying the agent.

Configuration LLM requise

Agents features have different LLM requirements. The table below shows which external LLM subscriptions are compatible with each feature.

Fonctionnalités	OpenAI	Google	Anthropic	Other open-source models
Exécution de l’agent	✅	✅	✅	✅
Agent evaluations	✅	✅	✅	✅
Ancrage dans le contexte	✅	✅	✅	Embeddings + Batch only 1
Autopilot for Agent Builder	❌	✅	✅	❌
Simulations	✅2	❌	❌	❌

¹ When using self-hosted models, Context Grounding only supports Embeddings and Batch Transform. DeepRAG and Advanced Extraction require a cloud model (OpenAI, Google, or Anthropic). ² Simulations require the GPT 4.1 mini model.

For more details, refer to Large Language Model Support and Configuring LLMs.