SmythOS Agents Integration

Want to build complex, multi-layered agents without reinventing the wheel? Connect SmythOS Agents to each other to create powerful, modular, and scalable automations.

TL;DR

Use the SmythOS Agent component to call one agent from within another (a "subagent"). This allows you to reuse logic, break down complex problems, and build sophisticated, hierarchical agent systems.

List of SmythOS Agent Components

The integration provides one powerful component for connecting agents together.

Component	Action	What it Does	Inputs	Key Outputs	Use Case
SmythOS Agent	Execute Agent	Runs another SmythOS agent as a sub-task and returns its result.	required prompt	Response	Creating a "manager" agent that delegates tasks.

INFO

Why Integrate Agents with Each Other?

Connecting agents is a core concept in building advanced AI systems. It allows you to move from single-purpose automations to complex, intelligent workflows.

• ♻️ Reusability: Build a specialized agent that does one thing well (e.g., summarize text, analyze sentiment). You can then reuse that "subagent" in dozens of other "parent" agents without duplicating any logic.
• 🧩 Modularity: Break down a large, complex problem into smaller, manageable sub-tasks. Assign each sub-task to a dedicated subagent. This makes your main agent easier to build, debug, and maintain.
• 📈 Scalability: Easily extend the capabilities of a parent agent by simply building a new subagent and calling it. This allows your systems to grow in complexity without becoming unmanageable.
• Hierarchical Control: Design "manager" agents that orchestrate multiple "worker" subagents. The manager can make high-level decisions and delegate the detailed work, mimicking a real-world team structure.

Prerequisites

Before you begin, please ensure you have the following:

• An active SmythOS account. (Sign up here).
• At least one existing SmythOS agent that can be used as a subagent (the agent being called).
• A second SmythOS agent to act as the parent agent (the agent that will do the calling).

Getting Started With Subagents

The process involves configuring a parent agent to call a subagent.

Step 1: Create Your Specialized Subagent

First, build and test an agent that performs a specific, reusable task. For example, create an agent named "Sentiment Analyzer" that takes a piece of text as input and outputs "Positive," "Negative," or "Neutral."

Step 2: Create Your Parent Agent

Next, create the agent that will orchestrate the work. For example, an agent named "Customer Feedback Processor."

Step 3: Add and Configure the SmythOS Agent Component

1. In your parent agent's graph ("Customer Feedback Processor"), drag and drop the SmythOS Agent component from the component library.
2. Click the component to open its Settings panel.
3. From the dropdown list, select the subagent you want to call (e.g., "Sentiment Analyzer").
4. Configure the Version settings to determine if you want to call the Dev or Prod version of the subagent.

No API Key Needed

Since you are connecting agents within the same SmythOS ecosystem, no external authentication like API keys is required. The connection is handled seamlessly.

Which Component Should I Use?

If you need to…	Use this Component	Why?
Make one agent perform the task of another agent	SmythOS Agent	This is the dedicated component for creating parent-child relationships between agents.

Component Details

This section provides detailed information for the SmythOS Agent component.

SmythOS Agent

Executes a specified "subagent" as part of the current agent's workflow, passing a prompt to it and receiving its final response.

INFO

This is an internal integration and does not require external API keys. Configuration is handled within the component's settings.

Component-Specific Settings

• Model: Choose the AI model the parent agent will use to reason about how and when to call the subagent.
• Version:
  • Same As Parent: If the parent is in Dev, it calls the Dev subagent. If Prod, it calls Prod.
  • Dev: Always calls the Dev version of the subagent, useful for testing.
  • Prod Latest: Always calls the latest Prod version of the subagent, ensuring stability.
• Description for Model: A description of what the subagent does. This is pre-filled from the subagent's Behavior field but can be edited. This text helps the parent agent's AI understand the subagent's capabilities.

Inputs

Field	Type	Required	Notes
prompt	string	Yes	The instruction or data passed from the parent agent to the subagent.

Outputs

Field	Type	Description
Response	any	The final output returned by the subagent. The data type depends on what the subagent is designed to produce.

Use Case

A "Manager Agent" receives an email. It uses an AI model to check if the email is a sales lead or a support ticket.

• If it's a sales lead, it calls the "Sales Lead Subagent" with the prompt containing the email body.
• If it's a support ticket, it calls the "Support Ticket Subagent" with the prompt containing the email body. The Manager Agent then takes the Response from the subagent (e.g., a CRM entry confirmation) and archives the email.

{
  "component": "smythos.agent",
  "prompt": "Please analyze the sentiment of the following customer feedback: 'The new feature is incredible and so easy to use!'"
}

Version Control is Crucial

The Version setting is critical. Using "Dev" allows for safe testing, while "Prod Latest" ensures your live automations are stable. Be mindful of which version you are targeting to avoid unexpected behavior.

Best Practices & Advanced Tips

• Design for Modularity: Build subagents that are highly specialized and do one thing exceptionally well. A good subagent is like a good function in programming: reliable, predictable, and reusable.
• Write Clear Descriptions: The "Description for Model" setting is vital. Write a clear, concise description of the subagent's purpose, inputs, and outputs. This helps the parent agent's AI make better decisions about when and how to use the subagent.
• Manage Versions: Thoroughly test your subagents in their Dev environment. Only promote a subagent to Prod when it is stable. In your parent agent, use the Dev version setting for testing and switch to Prod Latest for your live workflows.
• Standardize Inputs/Outputs: For easier interoperability, design your subagents to accept and return data in a consistent format, such as JSON.

Troubleshooting Common Issues

• Subagent Fails to Execute

  • Cause: The subagent itself might have an error, or the parent agent might be passing an improperly formatted prompt.
  • Solution: Check the execution history of the subagent directly. Its logs will contain the specific error message. Verify that the data being passed in the prompt from the parent matches what the subagent expects.

• Incorrect Output from Subagent

  • Cause: The prompt from the parent agent may be ambiguous, or the logic within the subagent may be flawed. You might also be calling the wrong version (Dev vs. Prod).
  • Solution: First, check the Version setting in the parent component. Ensure you are targeting the correct subagent version. Then, simplify the prompt to be as clear as possible. If the issue persists, debug the subagent's internal logic.

• Infinite Loops

  • Cause: Agent A calls Agent B, and Agent B calls Agent A, creating a recursive loop.
  • Solution: Carefully map out your agent architecture. Avoid circular dependencies where agents call each other in a way that could lead to an infinite loop.

What's Next?

You are now ready to build sophisticated, multi-agent systems with the SmythOS Agents Integration!

Consider these ideas:

• Build an Agent That...

  • Acts as a "Project Manager." It receives a high-level goal, breaks it down into tasks, and then calls different specialized subagents (a "Research Agent," a "Writing Agent," a "Publishing Agent") to complete each task in sequence.
  • Functions as a customer service dispatcher. It receives all incoming messages, uses a "Language Detection Subagent," a "Sentiment Analysis Subagent," and a "Topic Classification Subagent" to triage the message, and finally routes it to the correct human department.

• Explore Other Integrations:

  • A parent agent can fetch data from a Google Sheet, then use a Loop component to pass each row to a subagent for individual processing.
  • A subagent could be dedicated to a specific integration, like Ahrefs. This "Ahrefs Specialist" agent could be called by any parent agent that needs to perform SEO analysis.