Quick Start

Build a small ActionFlow: a Start node, Generate Text, and Slack. Studio autosaves. You run from the bottom bar, not from a Test or Save button.

What you will build

A graph that:

1. Starts on demand
2. Asks a model to analyze a short text
3. Posts the result to Slack

Step 1: Create the flow

1. Open ActionFlows.
2. Click New ActionFlow.
3. Name it, then click Create empty flow.

You land in Actionflow Studio. New flows start passive. Open the left Settings rail and click Set to active before you rely on webhooks, schedules, or dashboard runs. Studio can still edit a passive graph; inbound webhook POSTs return 409 while the flow is passive.

Step 2: Add Start

Open the start-node picker (or Start in Add node) and drop Start.

That is the on-demand entry. The other start nodes are Start Scheduled and Start Subflow. There is no Email, Webhook, or API start-node type.

Inbound HTTP uses a flow webhook on this same Start graph. Chat in Slack or email is an agent trigger, not a flow start node.

Start has no Inputs or Add Input panel. Empty required fields on downstream nodes become Flow inputs at run time ({nodeId}_{inputName}).

Step 3: Add Generate Text

1. Open Node Categories (plus on the left rail).
2. Under AI, add Generate Text.
3. Connect an execution edge from Start to Generate Text.
4. Select Generate Text. In the inspector, set Prompt, for example:

Analyze this text. Return:
1. Main topic
2. Sentiment (positive, negative, or neutral)
3. Three key takeaways

Text: JavaScript runs in the browser and on servers. Teams like its flexibility. Debugging async code can be hard.

5. Pick a model. If the organization uses its own provider keys, choose a named connection from Organization → Integrations → AI Vendors (or create one from the inspector).

Leave Prompt empty if you want to type the text in Flow inputs each run. The run form key is {generateTextNodeId}_prompt.

You can attach organization skills on the prompt (and system) fields. Skills are instruction text, not shell tools.

Step 4: Add Slack

1. In Node Categories, open Communication and add Slack.
2. Connect Generate Text to Slack.
3. Select Slack. Pick a named Slack connection (OAuth from Organization → Integrations, or Connect from the inspector).
4. Set the channel and message. Map the message body from the Generate Text output handle, or type a static message for the first run.

Step 5: Clear requirements and run

The bottom bar is how you execute:

Flow Requirements appears on that bar when a connected node is missing a credential. It lists missing org connections. It is not a linter for types, cost, or expiry dates.

If Prompt is filled and Slack is connected, click Start. If Prompt (or another required field) is empty, click Run flow, fill the form, then run.

Watch nodes on the canvas and open Run Output. Studio and dashboard runs use triggerSource manual. The same keys work on POST /api/runs and flow webhooks.

There is no Test button on the canvas. There is no isolated single-node playground.

Step 6: Run again from the list

Autosave writes the graph while you edit. To run later:

• Open the flow and use Start or Run flow again
• Or run it from the ActionFlows list (the flow must be active)

What you built

Start
  → Generate Text (model + prompt)
    → Slack (named org connection)

Next

• Webhook for inbound HTTP on the same Start graph
• Scheduled if the platform should fire the graph
• Actionflow Studio for the editor layout
• Credential management for named connections
• HITL Approvals to pause a run for a person

Troubleshooting

Start is disabled or the run fails validation

Confirm a start node exists, Generate Text has a model, required fields are filled or supplied as Flow inputs, and Slack has a named connection. Unconnected nodes are skipped by validateFlow.

Slack does not post

Reconnect Slack under Organization → Integrations. Confirm the bot can post to that channel.

Empty model output

Check the prompt, the selected model, and Run Output for the node error.

Webhook returns 409

The flow is still passive. Open Studio Settings and click Set to active.