Visual and Python Workflows

Modern EKB workflows come in two builder styles. You can create them yourself on the Tools canvas, or ask Autopilot (Build mode) to author them for you. This page covers how the two styles differ, how Autopilot fits in, and how to read the flowchart and summary views.

For the classic V1 vs V2 product lines and migration, see Workflows overview and Workflow Builder V2.

Visual vs Python at a glance

	Visual (Workflows v2)	Python (Workflows v3 / code mode)
What you edit	Drag-and-drop canvas of triggers, agents, tools, conditions, and outputs	Python source that defines the workflow logic
How you read it	Canvas graph	Readable flowchart with plain-English step labels, connected apps, and collapsible loop/branch cards — plus optional Developer (code) view
Best for	Processes you want to assemble and inspect visually	Logic that is easier to express in code, or that Autopilot should author end-to-end in Python
Autopilot in the editor	Autopilot helps via chat / create-with-AI flows	Autopilot chat stays docked on the canvas while you build and refine
Manual “New Flow” / blank tool	Default for manual creation	Not the default for blank manual create (see below)

A Python badge on a workflow in the Tools list means it is a code-mode (v3) workflow.

Creating a workflow

Build with Autopilot (recommended for either style)

1. Open your project (not the read-only Autopilot system project).
2. Use Autopilot in Build mode, or open Tools and start Build workflow with AI / the Autopilot create flow.
3. Describe what the workflow should do (you can attach files for context).
4. The create dialog shows whether Autopilot will build a Workflows v3 / Python or Workflows v2 / Visual workflow by default.
5. Autopilot plans, asks clarifying questions as needed, builds the workflow, and you can keep chatting to refine it.

ヒント

New Autopilot chats default to Agent mode. Switch to Build when you want Autopilot to create or change a workflow. See Autopilot Modes.

Who controls Visual vs Python for Autopilot creates?
A Super Admin setting — Default new Autopilot workflows to Workflows v3 — chooses the default:

• On — Autopilot creates Python / code mode workflows by default
• Off — Autopilot creates Visual (v2) workflows by default

Manual blank workflow creation stays on the Visual builder.

Create manually (Visual)

1. Go to Tools.
2. Use New Flow / create a blank custom tool.
3. Add a trigger, drag nodes onto the canvas, connect them, run, then publish.

Details: Workflow Builder V2.

Autopilot-built workflows

When Autopilot builds a workflow in Build mode it typically:

1. Clarifies the trigger, steps, apps, and outputs
2. Creates the workflow in Visual or Python mode (per the platform default above)
3. Configures steps / Python source
4. Validates and tests where applicable
5. Hands you a summary of what was created

You can continue in the same chat: refine steps, fix failures, or switch to Agent mode to run or reason about the result without further authoring. See How Autopilot Works.

Docked Autopilot in the editor

For Python (code mode) workflows, Autopilot stays docked on the editor canvas so you can describe changes while watching the flowchart and (in Developer view) the source update.

For Visual workflows, use Autopilot chat / “create with AI” flows to author and iterate; the canvas remains the place you drag and configure nodes.

Flowchart and summary views

Flowchart (Python workflows)

Python workflows render as a readable flowchart:

• Plain-English step labels (not only raw code identifiers)
• Connected apps called out on the flow
• Collapsible loop and branch cards for complex control flow
• Overview vs Developer detail: Overview emphasizes the business flow; Developer exposes the Python source alongside the graph

The same flowchart is used when inspecting a past run, with per-step status overlaid so you can see what executed, skipped, or failed.

Summary (Visual and Python)

Both Visual and Python workflows support a summary view — an at-a-glance description of what the workflow does, with structured steps. Summaries are generated/updated in connection with publishing; if you change the workflow after publish, the summary may need a fresh publish to stay current.

Super Admins can configure which model writes workflow summaries and step notes (platform model settings).

When to use which

Goal	Prefer
Assemble a process by dragging nodes and wiring conditions	Visual
Let Autopilot write substantial branching / data logic in code	Python
Keep Autopilot docked next to a live flowchart while iterating	Python
Follow existing V2 canvas docs and node catalog	Visual → Workflow Builder V2
“Build me a weekly report workflow” from chat	Autopilot Build mode (mode = platform default)

Related

• Workflows overview — V1 vs V2 product lines
• Workflow Builder V2 — Visual canvas how-to
• Autopilot overview — what Autopilot can build
• Autopilot Modes — Agent vs Build
• EKB as MCP — Builder — same build surface from Claude / Cursor
• 2.4.0 changelog — release note for Visual and Python Workflow Builder