Action Nodes

Action nodes do the actual work in your workflow. They’re where things happen: sending emails, asking AI to analyze something, adding rows to spreadsheets, posting to Slack, and more.

Available Action Nodes

Ask AI

Ask AI to write, summarize, analyze, categorize, or answer questions about text.

Call AI Agent

Run another workflow you’ve built (like calling a helper).

Execute Code

Run custom Python or JavaScript code for special tasks.

App Action

Do things in Gmail, Slack, Google Sheets, HubSpot, and hundreds of other apps.

HTTP Request

Connect to any website or service to send or receive data.

Update Variable

Save information to use in later steps.

Read File

Read text from PDFs, Word documents, images, and other files.

Which Action Should I Use?

I want to…	Use This
Have AI write, summarize, or analyze text	Ask AI
Use another workflow I already built	Call AI Agent
Do something custom with code	Execute Code
Send emails, post to Slack, update spreadsheets	App Action
Connect to a website or service not in the app list	HTTP Request
Save something to use later	Update Variable
Read text from a PDF, Word doc, or image	Read File
Have a back-and-forth conversation with someone	Conversation Flows

What Information Do Actions Provide?

Each action gives you information you can use in the next steps:

AI and File Actions

Node	What You Get
Ask AI	The AI’s response
Read File	The text from the file
Execute Code	Whatever your code returns

App and Web Actions

Node	What You Get
App Action	Depends on what you did (email ID, spreadsheet row, etc.)
HTTP Request	The response from the website/service

Saving Data

Node	What You Get
Update Variable	The value you saved
Call AI Agent	Whatever that workflow returned

What If Something Goes Wrong?

Sometimes things fail - a website is down, your email can’t send, etc. CogniAgent has ways to handle this:

Automatic Retries

Many actions (like Ask AI, HTTP Request, and App Action) automatically try again if they fail the first time. They’ll attempt up to 3 times before giving up.

Checking for Success

You can add a Condition node after an action to check if it worked:

├── HTTP Request (get customer data)
├── Condition (did it work?)
│   ├── Yes: Continue with the data
│   └── No: Send alert to Slack

Handling Errors in Code

If you’re using Execute Code, you can catch errors:

try:
    result = do_something_risky()
    return {"success": True, "data": result}
except Exception as e:
    return {"success": False, "error": str(e)}

Tips

For complex AI tasks, use multiple Ask AI nodes in sequence - one to analyze, one to decide, one to write. This makes it easier to see what’s happening at each step.

Use Update Variable to save results you’ll need later. It’s easier to use {{saved_result.value}} than to redo work.

Actions run one at a time in order, unless you’ve created separate paths with a Condition or Multi-Condition node.

Overview

Triggers

Actions

Logic

Communication

Available Action Nodes

Ask AI

Call AI Agent

Execute Code

App Action

HTTP Request

Update Variable

Read File

Which Action Should I Use?

What Information Do Actions Provide?

AI and File Actions

App and Web Actions

Saving Data

What If Something Goes Wrong?

Automatic Retries

Checking for Success

Handling Errors in Code

Tips

Next Steps

Ask AI Node

App Action

​Available Action Nodes

Ask AI

Call AI Agent

Execute Code

App Action

HTTP Request

Update Variable

Read File

​Which Action Should I Use?

​What Information Do Actions Provide?

​AI and File Actions

​App and Web Actions

​Saving Data

​What If Something Goes Wrong?

​Automatic Retries

​Checking for Success

​Handling Errors in Code

​Tips

​Next Steps

Ask AI Node

App Action

Available Action Nodes

Which Action Should I Use?

What Information Do Actions Provide?

AI and File Actions

App and Web Actions

Saving Data

What If Something Goes Wrong?

Automatic Retries

Checking for Success

Handling Errors in Code

Tips

Next Steps