Automation and AI
Workflow Definitions
Visual, node-based recipes you build in GreenVee's workflow editor, then deploy to run.
A Workflow Definition is a visual recipe for combining steps: a graph of nodes (individual steps) connected by edges (the order they run in). It defines what happens, in what order, every single time it runs. Unlike an Agent (which figures things out on the fly), a Workflow Definition is deterministic — step A always leads to step B, which always leads to step C.
Workflow Definition fields
Section titled “Workflow Definition fields”| Field | What it does |
|---|---|
| Name | A label for this workflow (e.g. “New Supplier Onboarding”) |
| Description | A note for your team — what this workflow is for |
| Status | Whether the workflow is draft, active, or archived |
| Graph | The nodes and connections you build in the workflow editor |
Agents vs. Workflow Definitions
Section titled “Agents vs. Workflow Definitions”Both Agents and Workflow Definitions can automate work. The difference is how they operate:
| Agent | Workflow Definition | |
|---|---|---|
| Approach | Reactive — given a task, it reasons and decides what to do | Deterministic — always follows the same fixed steps |
| Best for | Open-ended tasks, questions, triage, summarization | Structured processes that must run the same way every time |
| Flexibility | High — the Agent adapts | Low (by design) — no deviation from the recipe |
| Example | “Review this project update and flag any risks” | “When a supplier is added: send welcome email → create onboarding task → notify account manager” |
Use an Agent when the task requires judgment. Use a Workflow Definition when consistency and repeatability matter more than flexibility. The two aren’t exclusive: a workflow step can run an Agent, and an Agent can spawn a workflow-like subagent — see Workflow Nodes below.
Building a workflow
Section titled “Building a workflow”Open a Workflow Definition to get the visual editor. Use the node panel to search or browse for a step, drop it on the canvas, then draw connections between steps to set the order. Each node has its own settings form — what to do, and where its inputs come from (earlier steps’ outputs, the trigger’s data, or a fixed value).
Workflow Nodes: the building blocks
Section titled “Workflow Nodes: the building blocks”GreenVee ships 40+ node types you can combine, grouped into a few families:
| Family | What it covers | Examples |
|---|---|---|
| Flow control | The shape of the recipe itself | The entry-point trigger, a conditional branch (if/else), a loop over a list, a pause that waits for a signal, a child sub-workflow, the final result |
| AI | Steps that call a model or an Agent | An LLM step (ask a model a question), running an Agent, producing structured output |
| Data | Reading and writing GreenVee records | Create / update / delete a record, list / get records |
| Skills & Tools | Pulling in reusable capabilities mid-run | Load an Agent Skill, load MCP Tools, call an MCP Tool |
| Subagents | Delegating part of the work to an AI agent | Spawn a subagent, wait for a subagent to finish |
| Files | Working with files | Create a file, read a file |
| Notifications | Telling people what happened | Send a notification |
You don’t need to memorize the full list — the node panel lets you search by what you’re trying to do, and each node shows a short description of what it does.
Deploying
Section titled “Deploying”Editing a Workflow Definition never affects anything already running — GreenVee always executes a separate, frozen Deployment, not your live edits. When you’re ready to ship your changes, click Deploy:
- GreenVee runs a validation check over the whole graph (things like: does every input resolve, is there exactly one trigger, are database steps allowed to touch what they reference). Errors block the deploy; warnings don’t.
- If it passes, GreenVee freezes a new, numbered Deployment (v1, v2, v3, …) from the current graph.
- The header badge shows whether the graph matches the latest deployment, or has undeployed changes waiting.
Once deployed, the workflow starts running for real whenever its Trigger fires — see How Automation Runs for the full picture.
Running and watching a workflow
Section titled “Running and watching a workflow”If the workflow has a Manual trigger, you can start a run yourself with the Run button (only enabled once deployed). Every run — however it started — shows up live on the Workflow Executions page, where you can watch its progress step by step.
Related pages
Section titled “Related pages”- AI Glossary — plain-English definitions for Workflow Definition, Deployment, Trigger, and Execution
- Workflow Triggers — the four ways a workflow can start
- Workflow Executions — watching and reviewing runs
- How Automation Runs — the define → deploy → trigger → run → observe overview
- Agents — when to use an Agent instead of a Workflow Definition