Skip to main content
Tools are how an agent moves beyond conversation into action. A tool can look up account data, create a task, schedule an appointment, update a CRM record, or trigger a downstream workflow.

​
Tool Categories

  • Custom tools: HTTP requests you define
  • Pre-built tools: ready-made Feather capabilities exposed to agents
  • Integration-backed tools: actions unlocked after connecting a supported integration

​
When to Use a Tool

Use a tool when the agent needs trusted data or a side effect, for example:
  • Look up an order or policy
  • Create or update a CRM record
  • Trigger internal follow-up work
  • Book or confirm an appointment
  • Hand off to another workflow or human team

​
Custom Tool Example

This tool creates a follow-up task in an external system. 
curl -X POST "https://prod.featherhq.com/api/v1/tools" \
  -H "X-API-Key: $FEATHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Create follow-up task",
    "slug": "create_follow_up_task",
    "description": "Create a follow-up task in the CRM after a customer conversation.",
    "speech": "I am creating the follow-up now.",
    "structure": {
      "url": "https://api.example.com/tasks",
      "method": "POST",
      "headers": [
        {
          "key": "Authorization",
          "value": "Bearer YOUR_BACKEND_TOKEN",
          "secure": true
        },
        {
          "key": "Content-Type",
          "value": "application/json"
        }
      ],
      "body": {
        "title": "{{task_title}}",
        "ownerId": "{{owner_id}}",
        "customerId": "{{customer_id}}",
        "notes": "{{notes}}"
      },
      "timeout": 30,
      "variables": [
        {
          "name": "task_title",
          "type": "str",
          "description": "Short action-oriented task title",
          "required": true
        },
        {
          "name": "owner_id",
          "type": "str",
          "description": "CRM owner identifier",
          "required": true
        },
        {
          "name": "customer_id",
          "type": "str",
          "description": "Customer identifier in the CRM",
          "required": true
        },
        {
          "name": "notes",
          "type": "str",
          "description": "Freeform notes about the follow-up",
          "required": false
        }
      ]
    }
  }'

​
Prompting for Tool Use

Describe tools in the prompt in terms of intent, not implementation. Good prompts tell the agent:
  • When the tool should be used
  • Which facts must be confirmed first
  • What success looks like
  • When the agent must avoid using the tool

​
Tool Design Guidelines

  • Keep tools narrow and outcome-focused.
  • Require only the inputs the agent can reliably collect.
  • Return structured, human-readable results.
  • Hide credentials in secure headers instead of prompt text.
  • Make side effects idempotent where possible.

​
Tools and Knowledge Base

Tools and knowledge base solve different problems:
  • Use knowledge base for grounded answers from documents.
  • Use tools for live data retrieval and actions.
Most production agents use both.

Custom Tools API

Create and manage custom tools.

Pre-built Tools API

Inspect the tools Feather exposes directly.

Integrations

Learn how connected services can power tools.

Knowledge Base

Pair tools with grounded document retrieval.