Workflows
Build automated, multi-step processes on a visual node canvas — connect triggers, AI steps, control flow, and actions, then save, publish, run, and debug them. No code required, with code and HTTP escape hatches when you want them.
What is a workflow?
A workflow is a graph of connected nodes that runs automatically in response to a trigger. The trigger fires the workflow, AI nodes reason over data, control-flow nodes branch and loop, and action nodes do the work — call an API, run an integration, write to a knowledge folder, or notify your team. Nodes are wired together by arrows that define the order of execution.
Think "every time someone submits our intake form, analyse the request with AI, ask a manager to approve it, then create the record and notify the team" — turned into a reliable, observable, automated process that runs without anyone watching.
Workflows execute on a LangGraph runtime (a Python engine) behind a dedicated AI server — a fully managed automation and agent engine built into YOffice. The builder compiles your canvas into a graph; the runtime handles branching, loops, and native human-in-the-loop pauses. This is the same engine that runs Agents. You never see any of that plumbing; you just design on the canvas.
The Automate hub
Everything automation-related lives under Workflows in the navigation rail. The hub is organised into six tabs:
| Tab | What's there |
|---|---|
| Workflows | Your workflows — open one to edit on the canvas, or create a new one. |
| Templates | Pre-built workflows you can clone and customise. |
| Runs | Execution history and per-run logs for every workflow. |
| Forms | Fillable forms generated by form-trigger workflows. |
| Approvals | Pending human-approval steps waiting on a decision. |
| Assistants | Conversational, Assistant-trigger workflows — a visitor chats with one to kick off the run. |
The visual node-canvas builder
Open any workflow to launch the builder. You arrange nodes on a pannable canvas, wire them together by dragging from an output port to an input port, and configure each one in a side inspector. A node library lists every available node type, grouped by category. Changes autosave as a draft while you work, with a Saving / Saved indicator tracking the state.
The first node you add must always be a trigger — the builder only offers triggers on an empty canvas. From there you can add AI, logic, I/O, utility, and human-in-the-loop nodes in any arrangement. Building a workflow walks through the canvas in detail.
Design with Assistant
Don't want to wire nodes by hand? Open the embedded Design with Assistant chat inside the builder and describe what you want in plain language — "when a Stripe payment fails, draft an email to the customer and create a follow-up task". The assistant proposes and edits the graph for you: adding nodes, connecting them, and filling in their settings. You stay in control and can refine anything it builds directly on the canvas.
Build visually by dragging nodes, build conversationally by chatting, or mix the two — describe the skeleton with the assistant, then fine-tune individual nodes in the inspector. Both edit the same underlying graph.
Explore the workflow docs
This section breaks workflows down into focused guides. Start with building, then dig into triggers, the full node reference, running and approvals, and templates and forms.
AI nodes and integration actions draw on your organization's budget, and admins set per-service and per-workflow caps in Settings to keep costs predictable — when a budget is reached the relevant capability pauses automatically so you never overspend. Every numeric cap has a hard minimum, and 0 means disabled, never unlimited.