How to Create a Custom Tool in Langchain Easily

To create a custom tool in Langchain, define a class that inherits from BaseTool and implement the _call method with your tool's logic. Then, instantiate and use this tool within your Langchain agent or workflow.

Syntax

Creating a custom tool in Langchain involves subclassing BaseTool and overriding the _call method. The name and description attributes describe the tool for the agent.

• BaseTool: The base class for all tools.
• _call(self, input: str) -> str: The method where your tool's main logic goes.
• name: A string to identify the tool.
• description: A string explaining what the tool does.

python

from langchain.tools import BaseTool

class CustomTool(BaseTool):
    name = "custom_tool"
    description = "Describe what this tool does"

    def _call(self, input: str) -> str:
        # Your tool logic here
        return "Result based on " + input

Example

This example shows a custom tool that reverses the input string. It demonstrates how to define the tool and use it by calling its _call method.

python

from langchain.tools import BaseTool

class ReverseStringTool(BaseTool):
    name = "reverse_string"
    description = "Reverses the input string"

    def _call(self, input: str) -> str:
        return input[::-1]

# Using the custom tool
reverse_tool = ReverseStringTool()
output = reverse_tool._call("hello world")
print(output)

Output

dlrow olleh

Common Pitfalls

Common mistakes when creating custom tools include:

• Not overriding the _call method properly, which causes the tool to do nothing or error.
• Forgetting to set name and description, making the tool unclear to the agent.
• Using synchronous code when async is required in some Langchain contexts.

Always test your tool independently before integrating it into an agent.

python

from langchain.tools import BaseTool

# Wrong: Missing _call method
class BadTool(BaseTool):
    name = "bad_tool"
    description = "Does nothing"

# Right: Proper _call implementation
class GoodTool(BaseTool):
    name = "good_tool"
    description = "Returns input uppercased"

    def _call(self, input: str) -> str:
        return input.upper()

Quick Reference

Remember these key points when creating custom tools in Langchain:

• Subclass BaseTool.
• Implement the _call(self, input: str) -> str method.
• Set name and description for clarity.
• Test your tool standalone before integration.

Key Takeaways

Create custom tools by subclassing BaseTool and overriding the _call method.

Always provide clear name and description attributes for your tool.

Test your tool independently before using it in Langchain agents.

Avoid missing the _call method or incorrect method signatures.

Use synchronous or asynchronous code as required by your Langchain setup.