Overview

System tools are automatically available to every agent and don’t require custom configuration. These tools help agents orchestrate workflows, retrieve information, and provide better user experiences.

Available System Tools

RAG Tool

Retrieve information from datasources

Finish Tool

Complete current agent task

End Agent Graph Tool

Terminate the entire workflow

Agent Graph Insight Tool

Provide status updates to users

RAG Tool

Purpose

Retrieves relevant information from datasources attached to the agent block.

Usage

Simply mention in your agent prompt when the tool should be called:
Agent Prompt Example
The agent will automatically access the datasources you’ve attached to the agent block and retrieve the most relevant information.
Make sure to attach your datasources to the agent block before referencing the RAG Tool.

Finish Tool

Purpose

Signals that the current agent has completed its task and no further action is needed from this agent block.

Usage

Instruct the agent to call this tool when its specific objective is complete:
Agent Prompt Example

When to Use

  • Agent has gathered all required information
  • Task objective has been accomplished
  • Ready to pass control to the next agent or step
If your agent is finishing prematurely, explicitly define the conditions for calling FinishTool in your prompt.

End Agent Graph Tool

Purpose

Terminates the entire workflow immediately, regardless of remaining steps.

Usage

Use this tool when the entire flow should stop:
Agent Prompt Example

When to Use

  • User requests to cancel the process
  • Critical error that prevents continuation
  • Agent Graph objective has been fully completed
  • User wants to exit the conversation
This tool stops the entire workflow, not just the current agent. Use FinishTool if you only want to complete the current agent’s task.

Agent Graph Insight Tool

Purpose

Provides real-time status updates to users during tool execution, making conversations feel more natural and interactive.

Usage

The agent automatically uses this tool to inform users when processing is happening:
Agent Prompt Example

Example Responses

Benefits

  • Reduces user anxiety during processing time
  • Makes the conversation feel more human
  • Provides transparency about what’s happening
  • Improves overall user experience

Best Practices

If your agent is calling system tools at the wrong time:Do: Be explicit about when to call each tool
Don’t: Leave tool calling conditions ambiguous
If agents are ending flows too early:
  • Explicitly list all required steps before calling FinishTool/EndFlowTool
  • Add validation checkpoints in your prompt
  • Use conditional statements for tool calling
Example
  • Attach only relevant datasources to each agent
  • Provide clear instructions on when to search
  • Test retrieval quality with sample queries
  • Update datasources regularly for accuracy
Keep insight messages:
  • Brief and conversational
  • Relevant to the action being performed
  • Professional but friendly
  • Transparent about what’s happening

Troubleshooting

Add explicit conditions in your prompt for when to call FinishTool or EndFlowTool. List all required steps that must be completed first.
  • Verify datasources are properly attached to the agent block
  • Check that datasources contain the relevant information
  • Review the quality and format of your datasource content
  • Ensure the agent prompt mentions using FlowInsightTool
  • Verify the conversational flow is configured correctly
  • Check that the tool is being called before long-running operations