Getting Started
How ActionFlows is structured. Start nodes, Actionflow Studio, named org connections, and how a graph actually goes live.
Getting Started
ActionFlows is the platform. An ActionFlow is a directed graph you edit in Actionflow Studio. An Action Agent is a separate conversational assistant you design in Action Agent Studio.
How a flow run starts
A flow has exactly one start node:
| Start node | When it fires |
|---|---|
| Start | Studio Start / Run flow, dashboard run, POST /api/runs, or a flow webhook |
| Start Scheduled | Recurring, weekly, or one-time runAt |
| Start Subflow | A parent graph's Subflow node |
There is no Email, Webhook, or API start-node type. An inbound email does not start a flow by itself. Mailbox or helpdesk traffic should POST a flow webhook, or you run the graph with sample text in Flow inputs. Channel chat (Slack, Discord, Teams, email) is an agent trigger.
After the start node, you add AI, control, integration, storage, and HITL nodes. Wire execution and data handles. Nodes pick named organization connections. They do not store host, username, or password on the canvas.
How it goes live
Studio autosaves. Saving is not a deploy step.
New flows are created passive. Settings → Set to active is required before webhooks and schedules will fire. Passive flows cannot be run from the dashboard. A webhook POST to a passive flow returns 409.
A studio run executes the full current graph. There is no single-node playground.
Core objects
ActionFlow
A saved graph owned by an organization. See What is ActionFlow.
Node
One step. Catalog: Nodes. Configuration lives in the inspector. Empty required inputs become {nodeId}_{inputName} keys on Flow inputs, the API, and webhooks.
Start node
How the graph begins. See Manual, Scheduled, Subflow.
Action Agent
A durable chat assistant (prompt, model, skills, tools, MCP, triggers). See Action Agents.
Named connection
Org-scoped OAuth, API keys, databases, FTP, SMTP, S3, and MCP. See Credential management.
Run
One execution. Inspect it in Studio Run Output or the flow Flow Run History rail.
First graph, in product terms
- Create an ActionFlow (Create empty flow).
- Add Start.
- Add Generate Text (category AI) and an action such as Slack (category Communication).
- Pick a model and named connections.
- Set to active.
- Click Start, or Run flow if required fields are empty.
- Read Run Output. Change the prompt or wiring. Run again.
Follow the click-by-click version in Quick Start.
What to read next
- Actionflow Studio
- Trigger
- HITL Approvals
- Action Agents if you need multi-turn chat
- API Introduction to start runs from your backend
ActionFlows Documentation
Official ActionFlows documentation. Build, run, and automate AI-powered workflows with a visual canvas, triggers, integrations, SDKs, and a REST API.
Quick Start
Build and run your first ActionFlow in Actionflow Studio. Add a Start node, Generate Text, and Slack, then run from the bottom bar.