An Advanced agent is the full-featured agent type — it gets the complete configuration suite (instructions, tools, knowledge, memory, guardrails, and more) instead of the lighter Simple or Workflow forms.
Overview
When you create an agent and choose Advanced AI Agent — "AI to deliver personalized interactions across business workflows" — from the type picker, you get a short details form (avatar, name, description, department, workspace, tags), and then land in the full agent editor: a left side-nav of configuration sections next to a panel that opens whichever section you select. A new Advanced agent starts from your workspace's default settings — its indexing, response, pre-processing, and guardrail configuration are seeded from the global defaults, so it works before you customize anything.
How Advanced Compares to Simple and Workflow
Type | Editor | Best for |
|---|---|---|
Simple | A split-view form with a live test chat, no model selector in the header (model lives in a Settings drawer instead). | Standing up a useful agent quickly, without the full configuration suite. |
Workflow | A visual canvas of connected nodes (Branch, Condition, Path), started by a Form, Chat, or Schedule trigger. Available only in some environments. | Behavior that follows a defined, step-by-step process rather than open-ended conversation. |
Advanced | The full agent editor: a side-nav of every configuration section, with per-section save, Preview, Publish, and Version History. | Agents that need the complete configuration suite — tasks, memory, guardrails, multiple models, and fine-grained tools and knowledge. |
The Full Configuration Suite
What sets an Advanced agent apart is the range of sections in its editor's side-nav — each one configures a different part of how the agent behaves at runtime:
Section | Covers |
|---|---|
Instructions | The agent's role, goal, and behavior, plus the conversation starters users see first. |
Knowledge | The knowledge sets the agent retrieves from to answer accurately. |
Tools | Custom actions the agent can call: connector operations, MCP servers, custom endpoints, and workflows-as-tools. |
Skills | Reusable expertise attached to the agent. |
Capabilities | Built-in functions you switch on for the agent (web search, run code, create documents, and more). |
Tasks | Distinct journeys/intents the agent can run, each with its own objective, steps, and allowed tools. Permission-gated. |
Memory | What the agent remembers across conversations. |
Models | Which AI model runs each function of the agent. |
Guardrails | PII masking, hallucination checks, content filters, custom word filter, denied topics, and blocked messaging. Permission-gated. |
Settings | Indexing, pre-processing, response generation, and prerequisite actions. |
A Prompts library of reusable prompt templates is also available, reached from links inside the editor rather than the side-nav. Some sections only appear if your role grants access to them.
Tabs, Saving, and Publishing
The editor has three tabs — Configuration (everything above), Deployments, and Observability (the last two are permission-gated). Sections you type into show a per-section Save button; picker-based sections like Models and the Guardrails and Settings groups save automatically. In the header, Preview lets you try the agent and Publish makes your latest changes live (requires deploy permission), with Version History available once a version exists.
What to Configure Next
Because an Advanced agent's behavior is split across sections, the rest of the "Creating & Managing Agents" area and the sections that follow — Instructions, Tools, Knowledge, Memory, and Guardrails among them — each cover one section of this editor in depth. Start with Instructions to define who the agent is, then add the Tools, Skills, and Knowledge it needs to act, before tuning Memory and Guardrails.