Published at: 2026-09-17
AI PaaS quick start
This guide walks you through the shortest path to run the end-to-end minimal flow of AI PaaS: create an Agent → configure Skills → connect a knowledge base → start a conversation.
Completion Conditions
- An enterprise administrator has enabled Agent Studio and ShareAgent permissions in AI Platform Management.
- You have available knowledge base files or business object data (optional).
Step 1: Create an Agent in Agent Studio
Go to Setup → AI Platform Management → Agent Studio → Agent. In the basic information, fill in the name, API Name, and Agent settings (system prompt). Select the model provider and model version, then configure the welcome message and save.
For details, see 03-Agent Studio/02-Quick Start.md.
Step 2: Bind Skills and a knowledge base to the Agent
- From the Skills menu, select the Skills to mount from the enterprise skill library.
- From the Knowledge Base menu, bind an enterprise knowledge base or upload a local file. This lets the Agent retrieve internal enterprise documents.
For details, see 03-Agent Studio/03-Admin Guide.md.
Step 3: Verify the Agent in the debug preview area
In the debug preview area on the right side of the designer, enter a question. Check the Agent reply, reasoning content, and tool-chain call information. After you confirm the Agent behaves as expected, publish it as an in-use version.
Step 4: Invoke the Agent in ShareAgent
Users with permission can enter ShareAgent from the main navigation on the web. Create a conversation and enter a request (for example, “Generate an account brief for a customer”). ShareAgent recognizes the intent and discovers an available SubAgent to complete the request.
For details, see 04-ShareAgent/02-Quick Start.md.
Next steps
- Learn about billing: 07-Billing & Usage Management/01-Overview.md.
- Configure knowledge retrieval for an Agent: 06-Semantic Retrieval Index/02-Quick Start.md.