SaaS
How much do you leave to agents? Where do people review? How do you connect to your own systems? Once you know your way around the screens, this section shows you how to take AGENTIC STAR SaaS a step further.
Project
Bundle agents and steps, and carry out work while people make decisions at review steps. How to choose a type, where to place reviews, and the roles of playbooks and skills.
→API
Build SaaS agents into your own systems. Three connection methods (SSE, MCP, A2A) and REST APIs for conversations, files and artifacts.
→LLM Gateway
Just swap the base URL in your existing SDK. Hand model switching, cost caps, access control and usage visibility to the gateway.
→Admin
Decide your tenant's security in the admin console: how to decide sign-in and SSO, where people connect from and traffic going out, guardrails and records, and how they take effect.
→Getting Started
What is a Project
Learn how agents and steps are bundled, and how the two types differ.
→Call the API
Create a client in the Console and try the API.
→Call from the SDK
Swap the base URL of your SDK and call the gateway.
→Getting started with the admin console
What to set up, and in what order, before users start.
→Release Notes
The version history of the SaaS edition is in the Release Notes.
Two Kinds of API
If you need agent features such as tool execution, artifacts and conversation history, choose the API. If you only want to call a plain LLM in an OpenAI / Anthropic-compatible way, choose the LLM Gateway.
| Kind | Use | Main endpoint |
|---|---|---|
| API | Running autonomous agents (including tool execution and artifact generation) | /api/v1/chat/completions |
| LLM Gateway | Calling an LLM directly, OpenAI / Anthropic-compatible | /llm/v1/chat/completions |