Implementation:CrewAIInc CrewAI BaseTool Schema

Overview

Concrete abstract base class defining the tool interface with name, description, and Pydantic argument schema provided by the CrewAI framework.

Source Reference

Repository: crewAIInc/crewAI
File: lib/crewai/src/crewai/tools/base_tool.py
Lines: L55-96

Signature

class BaseTool(BaseModel, ABC):
    name: str
    description: str
    args_schema: type[BaseModel] = _ArgsSchemaPlaceholder
    result_as_answer: bool = False
    max_usage_count: int | None = None
    cache_function: Callable[..., bool]

    @abstractmethod
    def _run(self, **kwargs) -> str: ...

Fields

Field	Type	Default	Description
`name`	`str`	(required)	Unique identifier for the tool, used by the LLM to select it
`description`	`str`	(required)	Natural language description telling the LLM when and how to use the tool
`args_schema`	`type[BaseModel]`	`_ArgsSchemaPlaceholder`	Pydantic model defining accepted arguments with types and descriptions
`result_as_answer`	`bool`	`False`	If `True`, the tool's output is returned directly as the agent's final answer
`max_usage_count`	None	`None`	Maximum number of times this tool can be invoked per task execution; `None` means unlimited
`cache_function`	`Callable[..., bool]`	lambda returning True	Function that determines whether a tool result should be cached

Import

from crewai.tools import BaseTool

How It Works

BaseTool inherits from both Pydantic's BaseModel and Python's ABC (Abstract Base Class). This combination provides:

Schema validation via Pydantic -- all fields are type-checked and validated at instantiation time.
Abstract interface via ABC -- subclasses must implement the _run() method, which contains the actual tool logic.

The args_schema field defaults to an internal _ArgsSchemaPlaceholder class. When a subclass does not explicitly set args_schema, the framework introspects the _run() method's signature and generates a schema automatically. When a custom schema is provided, it takes precedence.

The framework serializes the tool specification (name, description, and args schema) into the format expected by the LLM's function-calling interface, enabling the LLM to select and invoke tools with correctly typed arguments.

Example

from pydantic import BaseModel, Field
from crewai.tools import BaseTool


class SearchArgs(BaseModel):
    """Arguments for the web search tool."""
    query: str = Field(description="The search query string")
    max_results: int = Field(default=5, description="Maximum number of results to return")


class WebSearchTool(BaseTool):
    name: str = "search_web"
    description: str = (
        "Search the web for current information about a topic. "
        "Use this when the agent needs up-to-date information "
        "that may not be in its training data."
    )
    args_schema: type[BaseModel] = SearchArgs

    def _run(self, query: str, max_results: int = 5) -> str:
        # Implementation of the actual search logic
        results = perform_search(query, max_results)
        return "\n".join(results)


# Instantiate and use the tool
search_tool = WebSearchTool()

In this example:

SearchArgs defines the typed argument schema with field descriptions.
WebSearchTool subclasses BaseTool and provides name, description, and args_schema.
The _run() method contains the actual tool logic.
The LLM sees the name, description, and schema when selecting tools.

Principle Link

Principle:CrewAIInc_CrewAI_Tool_Design

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment