How to beginner · 3 min read

LangChain tool calling agent

Q: LangChain tool calling agent

Use langchain to create a tool calling agent by defining tools with BaseTool and integrating them into an Agent with a language model like ChatOpenAI. The agent automatically decides when to call tools based on user input and returns combined results.

Quick answer

Use langchain to create a tool calling agent by defining tools with BaseTool and integrating them into an Agent with a language model like ChatOpenAI. The agent automatically decides when to call tools based on user input and returns combined results.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install langchain openai

Setup

Install the required packages and set your OpenAI API key as an environment variable.

Install LangChain and OpenAI SDK:

bash

pip install langchain openai

output

Collecting langchain
Collecting openai
Installing collected packages: ...
Successfully installed langchain-0.2.0 openai-1.0.0

Step by step

This example defines a simple tool that returns the current time, then creates a LangChain agent that can call this tool based on user queries.

python

import os
from langchain_openai import ChatOpenAI
from langchain.agents import Tool, initialize_agent, AgentType
from datetime import datetime

# Ensure your OpenAI API key is set in the environment
# export OPENAI_API_KEY=os.environ["OPENAI_API_KEY"]

# Define a simple tool
class TimeTool(Tool):
    name = "current_time"
    description = "Returns the current time as a string"

    def _run(self, query: str) -> str:
        return datetime.now().strftime("%Y-%m-%d %H:%M:%S")

    async def _arun(self, query: str) -> str:
        raise NotImplementedError("Async not implemented")

# Instantiate the tool
tools = [TimeTool()]

# Initialize the language model
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

# Create the agent with tool calling capability
agent = initialize_agent(
    tools,
    llm,
    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
    verbose=True
)

# Run the agent with a query that triggers the tool
response = agent.run("What time is it right now?")
print("Agent response:", response)

output

> Entering new AgentExecutor chain...
Thought: I need to use a tool to find the current time
Action: current_time
Action Input: None
Observation: 2026-04-27 15:42:10
Thought: I have the current time
Final Answer: The current time is 2026-04-27 15:42:10
> Finished chain.
Agent response: The current time is 2026-04-27 15:42:10

Common variations

Use async agent calls with await agent.arun() for asynchronous workflows.
Replace ChatOpenAI with other LLMs like Anthropic or OpenAI clients.
Define multiple tools for complex workflows, each with clear name and description.
Use different agent types like AgentType.CONVERSATIONAL_REACT_DESCRIPTION for chat-based tool calling.

Troubleshooting

If the agent does not call tools, ensure tools have proper name and description attributes.
Set verbose=True in initialize_agent to debug agent reasoning steps.
Check your OpenAI API key is correctly set in os.environ["OPENAI_API_KEY"].
For async errors, confirm your environment supports async and use await properly.

✅

Key Takeaways

Define tools as subclasses of Tool with clear names and descriptions for LangChain agents.
Use initialize_agent with AgentType.ZERO_SHOT_REACT_DESCRIPTION to enable automatic tool calling.
Set verbose=True to see agent decision-making and debug tool calls.
Async tool calling requires _arun method and async agent invocation.
Multiple tools can be combined for complex agent workflows.

Verified 2026-04 · gpt-4o-mini

Verify ↗