For AI agents: a documentation index is available at the root level at /llms.txt and /llms-full.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Atoms PlatformProduct OverviewDeveloper GuideAPI ReferenceMCPIntegrationsDeveloper ToolsChangelog
Atoms PlatformProduct OverviewDeveloper GuideAPI ReferenceMCPIntegrationsDeveloper ToolsChangelog
  • Get Started
    • Quickstart Crew CLI
    • Overview
    • Error Handling
  • Build
      • Overview
        • Overview
        • Defining Tools
        • Executing Tools
        • Built-in Tools
  • Operate
  • Examples
    • Examples
  • Migrate
    • From ElevenLabs
LogoLogo
Voice AgentsModels
Voice AgentsModels
On this page
  • What are Tools?
  • How It Works
  • Key Concepts
  • What’s Next
BuildAgent CrewsTools

Tools & Functions

||View as Markdown|
Was this page helpful?
Previous

Bring Your Own Model

Next

Defining Tools

Built with

Tools let your agent do things—look up orders, check weather, book appointments, or end a call. When a user asks something that requires real data or an action, the LLM decides to call a tool instead of making something up.

What are Tools?

Tools are Python functions that your agent can invoke during a conversation. The LLM analyzes the user’s request, decides which tool to call (if any), and your code executes it. The result is fed back to the LLM so it can respond naturally.

This is also known as function calling or tool use.

How It Works

  1. User says “What’s the weather in NYC?”
  2. LLM decides to call get_weather("New York, NY")
  3. Your function runs and returns {"temp": 72}
  4. LLM responds “It’s 72°F in New York City!”

The LLM never runs your code directly—it just tells you what to run and with what arguments.

Key Concepts

ConceptDescription
@function_tool()Decorator that marks a method as callable by the LLM
ToolRegistryDiscovers and manages all tools on your agent
Tool SchemasAuto-generated from your function’s docstring and type hints
Parallel ExecutionRun multiple tool calls at once

What’s Next

Defining Tools

The @function_tool decorator, docstrings, and type hints.

Executing Tools

ToolRegistry, handling calls, and feeding results back.

Built-in Actions

SDK-provided actions: ending calls, transferring to humans.