Skip to content
Nappai Documentation

Planner Executor Agent

The Planner Executor Agent is a smart tool within the Nappai automation system designed to solve complex problems that require more than just a simple command. Unlike standard tools that perform one specific action, this agent uses artificial intelligence to “think” about a task, create a step-by-step plan, and then carry out that plan to deliver a comprehensive result.

It is particularly useful when a task is difficult, involves multiple steps, or requires the agent to make decisions along the way. For example, if you ask it to “organize my data from last month,” it will first plan how to gather the data, then execute the steps to sort it, and finally present the organized result.

How it Works

This component operates using a “Planner-Executor” architecture. Here is a simple breakdown of its internal process:

  1. Planning Phase: When you provide a task or question, the agent first analyzes the request. It uses AI reasoning to break the problem down into smaller, manageable steps. It creates a logical “plan” of action rather than just jumping straight to an answer.
  2. Execution Phase: Once the plan is created, the agent executes each step in order. It may use various tools or look up information necessary to complete each step of the plan.
  3. Result Generation: After all steps are completed, the agent synthesizes the information and returns a final, coherent answer or result to you.

It is built on LangChain 1.0 technology, ensuring it follows modern standards for creating autonomous AI agents. The agent is designed to be self-correcting; if one step in the plan fails or encounters an issue, it will try to adjust its approach to still achieve the goal.

Connection & Credentials

This component does not require direct API credentials (such as external API keys) to be configured by the user in this view. However, it relies on the underlying system configuration where language model connections and general Nappai integrations are already set up. Ensure your general Nappai workspace has access to the necessary data sources and AI models for this agent to function correctly.

Inputs

This component primarily accepts the Task or Instruction that you want it to solve. The agent will take this input and process it through its planning and execution phases.

Outputs

The component produces a Final Result (usually text or a structured message). This output represents the completed task. Since the agent performs multiple steps internally, the output is the synthesized answer or the final state of the data after all planned actions have been executed.

Output Data Example (JSON)

json { “result”: “The data for Q3 has been successfully retrieved, filtered for ‘active’ status, and formatted into a summary table. The final report is attached below.”, “status”: “success”, “steps_executed”: 4 }

Connectivity

Typical Connections:

  • Preceding Components: This agent is typically placed at the start of a complex workflow or after a simple input node where you define the task (e.g., a “Input Text” or “User Prompt” node).
  • Following Components: The output of the Planner Executor Agent is usually connected to:
    • Output Nodes: To display the final result to the user in the dashboard.
    • Data Storage/Export Nodes: To save the result of the automated task into a database, spreadsheet, or document.
    • Other Agents: If the result of this plan needs further refinement, it can feed into another specialized agent.

Why it makes sense: Because this agent produces a comprehensive final answer, it serves as a powerful processing hub. It takes a complex query and transforms it into a usable result that other parts of your automation can rely on.

Usage Example

Scenario: You want to analyze customer feedback data from your CRM to identify the top 3 complaints.

  1. Input: You connect an Input Text component to the Planner Executor Agent. You type: “Analyze the last 100 customer reviews, identify the most common negative topics, and summarize the top 3 complaints.”
  2. Processing: The Planner Executor Agent receives this input.
    • Plan: It plans to first retrieve the reviews, then categorize them by topic, then count the frequency of negative mentions, and finally summarize the results.
    • Execute: It automatically performs these data retrieval and analysis steps.
  3. Output: The agent outputs a text summary: “Based on the analysis, the top 3 complaints are: 1. Slow shipping (30%), 2. Broken packaging (15%), 3. Poor customer service response time (10%).”
  4. Next Step: This result is then passed to a Chart Generator to create a visual report.

Important Notes

  • Development Status: This component is currently in the Development phase (is_development: True). This means it is experimental. The behavior or structure might change in future updates. It is recommended to test this component thoroughly before relying on it for critical, production-level automations.
  • Complexity Handling: Use this agent for tasks that are too complex for a simple direct tool. If a task can be done in one step, a simpler tool might be faster. Use this agent when you need the AI to “think” and break down the problem.
  • Icon: The component is identified by the AgentTodoIcon, which visually represents a task list or action sequence, helping you distinguish it from simple data tools.

Tips and Best Practices

  • Be Specific in Instructions: Since the agent plans your task, clear instructions lead to better plans. Instead of saying “Look at data,” say “Analyze the sales data from last month to find trends.”
  • Iterative Testing: Because the agent is in development, if you encounter unexpected behavior, try breaking down your request into simpler parts to see if the agent handles them better.
  • Monitor Outputs: Always review the output of the Planner Executor Agent to ensure the “plan” it followed aligns with your business logic, especially in critical workflows.

Security Considerations

  • Data Privacy: As this agent processes data, ensure that the data sources it connects to are secure and that you have permission to automate access to that information.
  • AI Hallucinations: AI agents can occasionally generate incorrect information. For critical decisions, use this agent to prepare drafts or summaries, but always verify the final output with human review if accuracy is paramount.