---
title: Agent execution
slug: agent-execution
docTags: 
createdAt: 2026-09-01T20:48:58.428Z
---

Users can access the agent deployment page to initiate various deployments, ensuring that agents operate with a configured language model, such as a hosted LLM or OpenAI LLM. This page allows users to set up all necessary configurations, including knowledge bases, policies, and human-in-the-loop (HITL) settings. When a user clicks on **Execute Deployment**, the system displays a dynamic form based on the input schema. 

The **Execute Deployment Page **displayed the following 3 options**: **

- **Basic Input Mode: **The form-generated payload aligns with the backend JSON schema, allowing the system to run the agent seamlessly in the chosen mode. When an agent requires ten fields, the dynamic form automatically generates them—such as text boxes or dropdown menus. This process streamlines data submission for both developers and users, reducing the likelihood of execution failures that can occur with incorrect payloads, which were more common with the previous manual JSON format.
- **Advanced Input Mode: **Add the raw JSON input manually. 
- **Upload Mode: **Users can upload a JSON file if they choose. Existing agents will continue to support JSON-based execution without changes.