Alpha Notice: These docs cover the v1-alpha release. Content is incomplete and subject to change.For the latest stable version, see the v0 LangChain Python or LangChain JavaScript docs.
Many AI applications interact with users via natural language. However, some use cases require models to interface directly with external systems—such as APIs, databases, or file systems—using structured input. Tools are components that agents call to perform actions. They extend model capabilities by letting them interact with the world through well-defined inputs and outputs. Tools encapsulate a callable function and its input schema. These can be passed to compatible chat models, allowing the model to decide whether to invoke a tool and with what arguments. In these scenarios, tool calling enables models to generate requests that conform to a specified input schema.

​
Create tools

​
Basic tool definition

The simplest way to create a tool is with the @tool decorator. By default, the function’s docstring becomes the tool’s description that helps the model understand when to use it: 
from langchain_core.tools import tool

@tool
def search_database(query: str, limit: int = 10) -> str:
    """Search the customer database for records matching the query.

    Args:
        query: Search terms to look for
        limit: Maximum number of results to return
    """
    return f"Found {limit} results for '{query}'"
Type hints are required as they define the tool’s input schema. The docstring should be informative and concise to help the model understand the tool’s purpose.

​
Customize tool properties

​
Custom tool name

By default, the tool name comes from the function name. Override it when you need something more descriptive: 
@tool("web_search")  # Custom name
def search(query: str) -> str:
    """Search the web for information."""
    return f"Results for: {query}"

print(search.name)  # web_search

​
Custom tool description

Override the auto-generated tool description for clearer model guidance: 
@tool("calculator", description="Performs arithmetic calculations. Use this for any math problems.")
def calc(expression: str) -> str:
    """Evaluate mathematical expressions."""
    return str(eval(expression))

​
Advanced schema definition

Define complex inputs with Pydantic models or JSON schemas: 
from pydantic import BaseModel, Field
from typing import Literal

class WeatherInput(BaseModel):
    """Input for weather queries."""
    location: str = Field(description="City name or coordinates")
    units: Literal["celsius", "fahrenheit"] = Field(
        default="celsius",
        description="Temperature unit preference"
    )
    include_forecast: bool = Field(
        default=False,
        description="Include 5-day forecast"
    )

@tool(args_schema=WeatherInput)
def get_weather(location: str, units: str = "celsius", include_forecast: bool = False) -> str:
    """Get current weather and optional forecast."""
    temp = 22 if units == "celsius" else 72
    result = f"Current weather in {location}: {temp} degrees {units[0].upper()}"
    if include_forecast:
        result += "\nNext 5 days: Sunny"
    return result

​
Use tools with agents

Agents go beyond simple tool binding by adding reasoning loops, state management, and multi-step execution.
To see examples of how to use tools with agents, see Agents.

​
Advanced tool patterns

​
ToolNode

ToolNode is a prebuilt LangGraph component that handles tool calls within an agent’s workflow. It works seamlessly with create_agent(), offering advanced tool execution control, built in parallelism, and error handling.

​
Configuration options

ToolNode accepts the following parameters: 
from langchain.agents import ToolNode

tool_node = ToolNode(
    tools=[...],              # List of tools or callables
    handle_tool_errors=True,  # Error handling configuration
    ...
)
​
tools
required
A list of tools that this node can execute. Can include:
  • LangChain @tool decorated functions
  • Callable objects (e.g. functions) with proper type hints and a docstring
​
handle_tool_errors
Controls how tool execution failures are handled. Can be:
  • bool
  • str
  • Callable[..., str]
  • type[Exception]
  • tuple[type[Exception], ...]
Default: internal _default_handle_tool_errors

​
Error handling strategies

ToolNode provides built-in error handling for tool execution through its handle_tool_errors property. To customize the error handling behavior, you can configure handle_tool_errors to either be a boolean, a string, a callable, an exception type, or a tuple of exception types:
  • True: Catch all errors and return a ToolMessage with the default error template containing the exception details.
  • str: Catch all errors and return a ToolMessage with this custom error message string.
  • type[Exception]: Only catch exceptions with the specified type and return the default error message for it.
  • tuple[type[Exception], ...]: Only catch exceptions with the specified types and return default error messages for them.
  • Callable[..., str]: Catch exceptions matching the callable’s signature and return the string result of calling it with the exception.
  • False: Disable error handling entirely, allowing exceptions to propagate.
handle_tool_errors defaults to a callable _default_handle_tool_errors that:
  • catches tool invocation errors ToolInvocationError (due to invalid arguments provided by the model) and returns a descriptive error message
  • ignores tool execution errors (they will be re-raised with the template string TOOL_CALL_ERROR_TEMPLATE = "Error: {error}\n Please fix your mistakes.")
Examples of how to use the different error handling strategies: 
# Retry on all exception types with the default error message template string
tool_node = ToolNode(tools=[my_tool], handle_tool_errors=True)

# Retry on all exception types with a custom message string
tool_node = ToolNode(
    tools=[my_tool],
    handle_tool_errors="I encountered an issue. Please try rephrasing your request."
)

# Retry on ValueError with a custom message, otherwise raise
def handle_errors(e: ValueError) -> str:
    return "Invalid input provided"

tool_node = ToolNode([my_tool], handle_tool_errors=handle_errors)

# Retry on ValueError and KeyError with the default error message template string, otherwise raise
tool_node = ToolNode(
    tools=[my_tool],
    handle_tool_errors=(ValueError, KeyError)
)

​
Use with create_agent()

We recommend that you familiarize yourself with create_agent() before covering this section. Read more about agents.
Pass a configured ToolNode directly to create_agent(): 
from langchain_openai import ChatOpenAI
from langchain.agents import ToolNode, create_agent
import random

@tool
def fetch_user_data(user_id: str) -> str:
    """Fetch user data from database."""
    if random.random() > 0.7:
        raise ConnectionError("Database connection timeout")
    return f"User {user_id}: John Doe, john@example.com, Active"

@tool
def process_transaction(amount: float, user_id: str) -> str:
    """Process a financial transaction."""
    if amount > 10000:
        raise ValueError(f"Amount {amount} exceeds maximum limit of 10000")
    return f"Processed ${amount} for user {user_id}"

def handle_errors(e: Exception) -> str:
    if isinstance(e, ConnectionError):
        return "The database is currently overloaded, but it is safe to retry. Please try again with the same parameters."
    elif isinstance(e, ValueError):
        return f"Error: {e}. Try to process the transaction in smaller amounts."
    return f"Error: {e}. Please try again."

tool_node = ToolNode(
    tools=[fetch_user_data, process_transaction],
    handle_tool_errors=handle_errors
)

agent = create_agent(
    model=ChatOpenAI(model="gpt-4o"),
    tools=tool_node,
    prompt="You are a financial assistant."
)

agent.invoke({
    "messages": [{"role": "user", "content": "Process a payment of 15000 dollars for user123. Generate a receipt email and address it to the user."}]
})
When you pass a ToolNode to create_agent(), the agent uses your exact configuration including error handling, custom names, and tags. This is useful when you need fine-grained control over tool execution behavior.

​
State, context, and memory

state: The agent maintains state throughout its execution - this includes messages, custom fields, and any data your tools need to track. State flows through the graph and can be accessed and modified by tools.
InjectedState: An annotation that allows tools to access the current graph state without exposing it to the LLM. This lets tools read information like message history or custom state fields while keeping the tool’s schema simple.
Tools can access the current graph state using the InjectedState annotation:
from typing_extensions import Annotated
from langchain.agents.tool_node import InjectedState

# Access the current conversation state
@tool
def summarize_conversation(
    state: Annotated[dict, InjectedState]
) -> str:
    """Summarize the conversation so far."""
    messages = state["messages"]

    human_msgs = sum(1 for m in messages if m.__class__.__name__ == "HumanMessage")
    ai_msgs = sum(1 for m in messages if m.__class__.__name__ == "AIMessage")
    tool_msgs = sum(1 for m in messages if m.__class__.__name__ == "ToolMessage")

    return f"Conversation has {human_msgs} user messages, {ai_msgs} AI responses, and {tool_msgs} tool results"

# Access custom state fields
@tool
def get_user_preference(
    pref_name: str,
    preferences: Annotated[dict, InjectedState("user_preferences")]  # InjectedState parameters are not visible to the model
) -> str:
    """Get a user preference value."""
    return preferences.get(pref_name, "Not set")
State-injected arguments are hidden from the model. For the example above, the model only sees pref_name in the tool schema - preferences is not included in the request.
Command: A special return type that tools can use to update the agent’s state or control the graph’s execution flow. Instead of just returning data, tools can return Commands to modify state or direct the agent to specific nodes.
Use a tool that returns a Command to update the agent state:
from langgraph.types import Command
from langchain_core.messages import RemoveMessage
from langgraph.graph.message import REMOVE_ALL_MESSAGES
from langchain_core.tools import tool, InjectedToolCallId
from typing_extensions import Annotated

# Update the conversation history by removing all messages
@tool
def clear_conversation() -> Command:
    """Clear the conversation history."""

    return Command(
        update={
            "messages": [RemoveMessage(id=REMOVE_ALL_MESSAGES)],
        }
    )

# Update the user_name in the agent state
@tool
def update_user_name(
    new_name: str,
    tool_call_id: Annotated[dict, InjectedToolCallId]
) -> Command:
    """Update the user's name."""
    return Command(update={"user_name": new_name})
runtime: The execution environment of your agent, containing immutable configuration and contextual data that persists throughout the agent’s execution (e.g., user IDs, session details, or application-specific configuration).
Tools can access an agent’s runtime context through get_runtime:
from dataclasses import dataclass
from langchain_openai import ChatOpenAI
from langchain.agents import create_agent
from langchain_core.tools import tool
from langgraph.runtime import get_runtime

USER_DATABASE = {
    "user123": {
        "name": "Alice Johnson",
        "account_type": "Premium",
        "balance": 5000,
        "email": "alice@example.com"
    },
    "user456": {
        "name": "Bob Smith",
        "account_type": "Standard",
        "balance": 1200,
        "email": "bob@example.com"
    }
}

@dataclass
class UserContext:
    user_id: str

@tool
def get_account_info() -> str:
    """Get the current user's account information."""
    runtime = get_runtime(UserContext)
    user_id = runtime.context.user_id

    if user_id in USER_DATABASE:
        user = USER_DATABASE[user_id]
        return f"Account holder: {user['name']}\nType: {user['account_type']}\nBalance: ${user['balance']}"
    return "User not found"

model = ChatOpenAI(model="gpt-4o")
agent = create_agent(
    model,
    tools=[get_account_info],
    context_schema=UserContext,
    prompt="You are a financial assistant."
)

result = agent.invoke(
    {"messages": [{"role": "user", "content": "What's my current balance?"}]},
    context=UserContext(user_id="user123")
)
store: LangChain’s persistence layer. An agent’s long-term memory store, e.g. user-specific or application-specific data stored across conversations.
Tools can access an agent’s store through get_store:
from langgraph.config import get_store

@tool
def get_user_info(user_id: str) -> str:
    """Look up user info."""
    store = get_store()
    user_info = store.get(("users",), user_id)
    return str(user_info.value) if user_info else "Unknown user"
To update long-term memory, you can use the .put() method of InMemoryStore. A complete example of persistent memory across sessions:
from typing import Any
from langgraph.config import get_store
from langgraph.store.memory import InMemoryStore
from langchain.agents import create_agent
from langchain_core.tools import tool

@tool
def get_user_info(user_id: str) -> str:
    """Look up user info."""
    store = get_store()
    user_info = store.get(("users",), user_id)
    return str(user_info.value) if user_info else "Unknown user"

@tool
def save_user_info(user_id: str, user_info: dict[str, Any]) -> str:
    """Save user info."""
    store = get_store()
    store.put(("users",), user_id, user_info)
    return "Successfully saved user info."

store = InMemoryStore()
agent = create_agent(
    model,
    tools=[get_user_info, save_user_info],
    store=store
)

# First session: save user info
agent.invoke({
    "messages": [{"role": "user", "content": "Save the following user: userid: abc123, name: Foo, age: 25, email: foo@langchain.dev"}]
})

# Second session: get user info
agent.invoke({
    "messages": [{"role": "user", "content": "Get user info for user with id 'abc123'"}]
})
# Here is the user info for user with ID "abc123":
# - Name: Foo
# - Age: 25
# - Email: foo@langchain.dev