A Simple agent is a lightweight AI agent you build in a split-view editor — a configuration form on the left and a live chat to test it on the right — for standing up a useful agent fast, without the full configuration suite of an Advanced agent.
Overview
The Simple agent editor's split view pairs the configuration form with a live Test Agent chat, so you can try the agent as you build it. When the AI agent builder is enabled for your workspace, you can build it two ways, switched with a header toggle: AI Assisted — describe what you want and a copilot conversation builds the agent for you — or Manual configuration, the form described below; new agents open in AI Assisted. Otherwise there's no toggle and the editor opens directly in Manual configuration.
The editor has three tabs: Configuration, Deployments, and Observability (the last two appear only if your role can deploy or view observability). The header carries the agent's name and avatar, a preview toggle, Share and Settings, and a single Save button — saving also publishes your changes in one step, so there is no separate Publish button. Save is disabled when there's nothing new to save or publish. There's no model selector in the header: the agent's model is chosen in the Settings drawer.
Configuration Sections
The Configuration form has a handful of sections, in order:
Section | Description |
|---|---|
Instructions | "Define your agent's role, tone, and behaviour." |
Knowledge sources | "Upload files the agent can reference when responding." Add Knowledge opens a three-item menu: Upload Documents (max file size 20 MB — "PDF, PPTX, DOCX and more"), Add Knowledge Set (shown only if you can create knowledge sets), and Connect an App. |
Tools | "Connect tools your agent can use — like sending emails or updating records." Added with Add Tools; these are your connected-app actions. Built-in capabilities you've switched on are not listed here — turn them off from the Capabilities section, not from this list. |
Skills | "Extend what your agent can do with ready-made processing skills." Added with Add Skills from the skills library. |
Capabilities | "Enable built-in capabilities like web search or image generation." Add Capabilities opens the built-in capabilities library (for example Web Search, Deep Research, Execute Code, Generate Image, Generate Chart, and Create DOCX / PDF). Once you've added more than three, the section shows the first three and a "+ N more" link. |
Conversation Starters | Up to 4 suggested prompts ("Add up to 4 prompts to help users get started quickly"). |
As you edit, the Test Agent panel on the right lets you message the agent immediately.
Note: Unlike the Advanced agent's Capabilities tab, the Simple editor's Capabilities section is a plain add/remove list — there's no per-capability Configure button or model picker here, so a capability like Web Search always runs on its provider's default model.
Settings Drawer
A Settings drawer holds the agent's avatar, name, description, Department (if your workspace uses departments), Model, Tags, Output Format (Text, or Object with a response schema), and Chat Input Placeholder. This is the only place a Simple agent's model is chosen — there is no model selector in the header.
Tools, Skills, and Tasks
On an agent these are related but distinct:
A tool is an action the agent can take — a connector operation, a workflow exposed as a tool, or one of the built-in capabilities (like web search, which the Simple editor lists in its own Capabilities section). The agent calls a tool when it needs to do something.
A skill is reusable expertise — "Skills are reusable expertise that teach your agent how to think and respond." Skills come from a shared skill library and shape how the agent reasons.
A task (on Advanced agents) is a whole journey/intent the agent can run, with its own objective, steps, and allowed tools — this concept doesn't appear on the Simple editor.
Tools and skills are added from their own sections; you add a skill from the skill library, and skills are linked to the agent when you save.
Knowledge Sync Status
In the Simple agent editor's Knowledge section, each source you add shows a status badge next to it:
Badge | Meaning |
|---|---|
Uploading… | The file is still transferring. |
Syncing… (with a spinner) | The source is being indexed; a just-added or queued source shows the same badge. |
Sync Error | Indexing failed; hover the row for a Retry action. |
Synced: <date> (or Completed) | The source is indexed and searchable. |
While a source is syncing the agent still works, but that source's newly added content isn't searchable yet; you may also see a dismissible note on the preview pane that knowledge is syncing in the background — the per-source badge is the reliable indicator. Once the badge shows Synced, test questions that rely on the new content.
Note: Knowledge sets attached to the agent appear as plain rows without a sync badge — their status lives on the knowledge set itself. The fuller per-source detail — sources grouped by type, each with an Overview page showing Crawled / Pending / Synced / Failed — lives on a standalone Knowledge Set, the surface Advanced agents use.