Skip to main content
Tools are the actions your agents can perform — from searching knowledge to updating tickets to calling external APIs.

​
Tool Categories

​
Duckie Tools (Built-in)

Core capabilities that come with every agent:
ToolDescriptionWhen It’s Used
ResponderGenerates and sends responses to customersEvery conversation that needs a reply
Ask and WaitAsks clarifying questions and waits for replyWhen more information is needed
Knowledge SearchSearches your knowledge baseFinding information to answer questions
EscalatorHands off to human agentsWhen guardrails trigger or agent can’t help
LLM CallCustom AI operationsData transformation, summarization, analysis
Save ValueStores values for later stepsPassing data between workflow nodes
End RunExplicitly ends the agent runTerminating workflows

​
App Tools (Integrations)

Tools from connected platforms — automatically available when you connect an integration:
PlatformExample Tools
SlackSend message, Reply in thread, Add reaction, Create channel
ZendeskUpdate ticket, Add internal note, Change status, Add tags
JiraCreate issue, Add comment, Update fields, Transition status
LinearCreate issue, Update issue, Add comment
IntercomSend message, Add note, Update conversation
HubSpotCreate ticket, Update contact, Add note

​
Custom Tools

Your own HTTP-based tools for org-specific integrations:
  • Any REST API — Connect to internal services or third-party APIs
  • Authentication — Bearer tokens, Basic auth, API keys
  • Parameter Mapping — Dynamic values from conversation context
  • Full Control — Define method, headers, body, and response handling

​
Using Tools

​
In Runbooks

Reference tools inline in your instructions: 
Ask for the customer's order number.

Use **[Order Lookup]** to find their order in the system.

If the order is found, share the status and tracking information.
The agent knows to call that specific tool when following these instructions.

​
In Workflows

Add Action nodes that execute tools:
  1. Add an Action node
  2. Select the tool to execute
  3. Map inputs from conversation or previous nodes
  4. Handle success and failure outputs

​
Creating Custom Tools

1

Navigate to Tools

Go to Build → Tools in your dashboard.
2

Click Create Tool

Click Create Custom Tool.
3

Configure Endpoint

Enter the API details:
  • URL — The endpoint to call (can include parameters)
  • Method — GET, POST, PUT, DELETE, etc.
  • Headers — Any required headers
4

Set Up Authentication

Choose authentication method:
  • None — No authentication
  • Bearer Token — Authorization header with token
  • Basic Auth — Username and password
  • API Key — Key in header or query parameter
5

Define Parameters

Specify what inputs the tool accepts:
{
  "order_id": {
    "type": "string",
    "description": "The order ID to look up",
    "required": true
  }
}
6

Configure Response

Define how to handle the response:
  • Which fields to extract
  • How to format for the agent
7

Test

Test with sample inputs to verify it works.

​
Parameter Placeholders

Use {{param}} syntax in URLs and request bodies: URL with parameter: 
https://api.example.com/orders/{{order_id}}
Request body: 
{
  "customer_email": "{{email}}",
  "action": "refund",
  "amount": "{{refund_amount}}"
}
Parameters are filled from:
  • Conversation context
  • Values extracted by previous steps
  • Static configuration

​
Tool Permissions

Control which tools each agent can use:
  1. Open agent configuration
  2. Go to Tools tab
  3. Enable or disable specific tools
This lets you:
  • Restrict sensitive actions to specific agents
  • Create agents with limited capabilities
  • Prevent unintended tool usage

​
Duckie Tools Reference

​
Responder

Generates and sends a response to the customer. When used: Automatically when the agent needs to reply. What it does:
  1. Gathers context from knowledge search
  2. Applies guidelines for tone and format
  3. Generates the response
  4. Sends through the source channel

​
Ask and Wait

Asks a clarifying question and pauses until the customer replies. When used: When the agent needs more information. What it does:
  1. Sends the clarifying question
  2. Pauses execution
  3. Resumes when customer responds
  4. Continues with the new information
Searches your knowledge base for relevant information. When used: Before generating responses, or explicitly in runbooks. Modes:
  • Basic — Simple semantic search
  • Advanced — Multi-query search with reranking

​
Escalator

Hands off the conversation to a human agent. When used: When guardrails trigger or agent cannot help. What it does:
  1. Sends a message to the customer
  2. Creates an internal note with context
  3. Routes to human queue
  4. Ends the agent run

​
LLM Call

Makes a custom AI call for specialized processing. Use cases:
  • Data transformation
  • Summarization
  • Sentiment analysis
  • Custom classification

​
Save Value

Stores a value for use in later workflow steps. Use case: Passing data between nodes in a workflow.

​
End Run

Explicitly terminates the agent run. Use case: Ending a workflow at a specific point.

​
Next Steps

Runbooks

Use tools in runbooks

Workflows

Use tools in workflow actions

Integrations

See available app tools