How to Use Action Nodes to Run AI Agents
This guide explains how to choose and use the correct Action nodes in CogniAgent to execute AI-powered workflows. Since you’ve already learned about the starting nodes, it’s time to master the most important steps that actually make things happen in your automation.
A Quick Refresher: What Is CogniAgent?
CogniAgent is an Event-Driven Automation (EDA) platform.
In EDA, every node in your workflow represents an event or a reaction to an event. Instead of manually triggering tasks, systems automatically respond to real-time signals such as:
- A new email arriving
- A user submitting a form
- A database record being updated
- A webhook or API event firing
These events trigger predefined actions – removing manual work from processes like incident handling, data processing, notifications, or AI-powered analysis.
Think of a workflow as a chain of intelligent reactions:
“When this happens → do that → decide → repeat or stop.”
What Are Action Nodes?
Action nodes are the heart of your workflow.
They define what the system actually does after an event occurs.
In simple terms:
“Once the workflow starts, go here and act like this.”
CogniAgent provides several types of Action nodes, each designed for a specific type of execution.
Call Agent Node
The Call Agent node allows you to invoke an AI agent inside your workflow.
Before using this node, you need to create your agents. CogniAgent supports:
- Autonomous agents – operate independently and execute tasks
- Conversational agents – interact with users or other systems via dialogue
Example use cases:
- Run a sentiment analysis agent on incoming customer feedback
- Trigger a research agent to summarize competitor updates
- Use a conversational agent to reply to inbound support requests
Once created, these agents can be reused across multiple workflows.
HTTP Request Node
The HTTP Node allows your workflow to securely communicate with external systems using HTTP requests.
It exists as a separate node because the Python code sandbox in CogniAgent is intentionally restricted:
- No access to the internal network
- No HTTP calls to webhooks of other workflows
- Limited external network access
These restrictions protect the platform from security risks and unintended workflow chaining.
Examples:
Use the HTTP Node whenever your workflow needs to:
- Call external APIs
- Send or receive data via webhooks
- Trigger another CogniAgent workflow
- Integrate with third-party services
Action in App Node
The Action in App node lets your workflow interact with third-party applications.
It allows you to send data to, or perform actions inside, external systems such as:
- CRM tools
- Messaging platforms
- Databases
- Project management tools
Examples:
- Send a Slack or Telegram message
- Create or update a CRM record
- Post content to a social media platform
- Add a task to a project board
Think of this node as a digital assistant that executes instructions inside other apps on your behalf.
Execute Code Node
The Execute Code node provides a powerful scripting environment for advanced logic.
It allows you to run custom Python code when:
- Built-in nodes are not flexible enough
- You need complex transformations or calculations
- You want full control over data handling
Common use cases:
- Advanced data filtering or restructuring
- Custom scoring algorithms
- Financial calculations or statistical analysis
- Parsing and normalizing complex JSON payloads
Although basic Python knowledge is recommended, CogniAgent makes this easier by allowing you to:
- Describe what you want in natural language
- Let an AI assistant generate the Python code for you
This makes the node accessible even to non-developers.
Set Variable Node
The Set Variable node allows you to create, store, and update data points that can be reused anywhere downstream in your workflow.
Unlike standard nodes – which only pass data to the next step – variables persist and remain accessible throughout the workflow.
When should you use it?
- To track whether a condition was met earlier
- To store intermediate results
- To control branching logic later in the flow
Variable integration allows you to reference workflow data dynamically inside your code or logic.
You can think of it in simple terms:
“If this event happens or meets this specific expectation – remember it – and act on it.”
Variables are especially useful:
- To track whether a condition was met earlier
- To store intermediate results
- To control branching logic later in the flow
LLM Nodes (Large Language Models)
LLM nodes enable text-based AI processing using models such as:
- GPT-4
- Claude
- Gemini
- NanoBanana
- Other.
This node acts as a bridge between your workflow data and an AI “brain.”
It:
- Takes structured input and instructions
- Sends them to the selected language model
- Returns a refined, human-like response
Example use cases:
- Summarizing documents or emails
- Generating replies or reports
- Classifying content or extracting insights
- Rewriting text in a specific tone or format
- Writing posts to your social channels, etc.
Parse File Node in CogniAgent
The Parse File node allows your workflow to read, extract, and structure data from uploaded files so it can be used by AI agents, logic nodes, or integrations downstream.
In simple terms:
This node turns files into usable data. Instead of treating a file as an opaque attachment, the Parse File node converts its contents into structured information that your workflow can understand, analyze, and act on.
When Should You Use the Parse File Node?
Use the Parse File node whenever your workflow needs to work with the contents of a file, not just store or forward it.
Typical scenarios include:
- Analyzing documents with AI
- Extracting text or data for processing
- Validating or classifying file content
- Extracting images, audio and video files → Feeding file data into LLMs or agents
Depending on your CogniAgent setup, the Parse File node supports documents, images, audio, and video files. It accepts a file URL or CogniAgent file descriptor.
The node automatically extracts the readable content and converts it into a structured output format.
Conclusion
Action nodes are what transform a CogniAgent workflow from a simple trigger into a living, intelligent system.
They allow you to:
- Activate AI agents
- Execute logic and calculations
- Interact with external applications
- Store memory across the workflow
- Understand and process documents
- Generate human-like language with LLMs
In short, Action nodes define what happens once an event occurs. They are responsible for execution, intelligence, and interaction – turning raw signals into meaningful outcomes.
However, execution alone is not enough. To build workflows that are truly autonomous, adaptive, and production-ready, you also need a way to decide, branch, repeat, or stop actions based on data and outcomes. That’s where Control Nodes come in.
What’s Next? Learn How to Control Your Workflows
In the next guide, we’ll cover Control Nodes – the final building blocks that determine how your workflow behaves under different conditions.
You’ll learn:
- How to route workflows using Conditions
- How to process data in bulk with Loops
- How to build decision-driven logic instead of linear flows
- How to combine Action and Control nodes for scalable automation
👉 Continue to the next guide: How to Use Control Nodes in CogniAgent
