custom_tool() — langchain Function Reference

Architecture documentation for the custom_tool() function in custom_tool.py from the langchain codebase.

Function python LangChainCore MessageInterface calls 2

Entity Profile

LangChainCore→ MessageInterface→ custom_tool() — langchain Function Reference

Dependency Diagram

graph TD
  36d50a89_56e5_c281_6ef1_1f4a9574ce0a["custom_tool()"]
  2bc1c3fe_4564_705b_4b4f_ef73fdf56b3c["custom_tool.py"]
  36d50a89_56e5_c281_6ef1_1f4a9574ce0a -->|defined in| 2bc1c3fe_4564_705b_4b4f_ef73fdf56b3c
  96434308_aeec_dce2_9fb2_a803ae096f07["_make_wrapped_coroutine()"]
  36d50a89_56e5_c281_6ef1_1f4a9574ce0a -->|calls| 96434308_aeec_dce2_9fb2_a803ae096f07
  f0fb9f2e_4f8a_b540_b097_0a1c095ba3a2["_make_wrapped_func()"]
  36d50a89_56e5_c281_6ef1_1f4a9574ce0a -->|calls| f0fb9f2e_4f8a_b540_b097_0a1c095ba3a2
  style 36d50a89_56e5_c281_6ef1_1f4a9574ce0a fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

libs/partners/openai/langchain_openai/tools/custom_tool.py lines 27–111

def custom_tool(*args: Any, **kwargs: Any) -> Any:
    """Decorator to create an OpenAI custom tool.

    Custom tools allow for tools with (potentially long) freeform string inputs.

    See below for an example using LangGraph:

    ```python
    @custom_tool
    def execute_code(code: str) -> str:
        \"\"\"Execute python code.\"\"\"
        return "27"


    model = ChatOpenAI(model="gpt-5", output_version="responses/v1")

    agent = create_react_agent(model, [execute_code])

    input_message = {"role": "user", "content": "Use the tool to calculate 3^3."}
    for step in agent.stream(
        {"messages": [input_message]},
        stream_mode="values",
    ):
        step["messages"][-1].pretty_print()
    ```

    You can also specify a format for a corresponding context-free grammar using the
    `format` kwarg:

    ```python
    from langchain_openai import ChatOpenAI, custom_tool
    from langgraph.prebuilt import create_react_agent

    grammar = \"\"\"
    start: expr
    expr: term (SP ADD SP term)* -> add
    | term
    term: factor (SP MUL SP factor)* -> mul
    | factor
    factor: INT
    SP: " "
    ADD: "+"
    MUL: "*"
    %import common.INT
    \"\"\"

    format = {"type": "grammar", "syntax": "lark", "definition": grammar}

    # highlight-next-line
    @custom_tool(format=format)
    def do_math(input_string: str) -> str:
        \"\"\"Do a mathematical operation.\"\"\"
        return "27"


    model = ChatOpenAI(model="gpt-5", output_version="responses/v1")

    agent = create_react_agent(model, [do_math])

    input_message = {"role": "user", "content": "Use the tool to calculate 3^3."}
    for step in agent.stream(
        {"messages": [input_message]},
        stream_mode="values",
    ):
        step["messages"][-1].pretty_print()
    ```
    """

    def decorator(func: Callable[..., Any]) -> Any:
        metadata = {"type": "custom_tool"}
        if "format" in kwargs:
            metadata["format"] = kwargs.pop("format")
        tool_obj = tool(infer_schema=False, **kwargs)(func)
        tool_obj.metadata = metadata
        tool_obj.description = func.__doc__
        if inspect.iscoroutinefunction(func):
            tool_obj.coroutine = _make_wrapped_coroutine(func)
        else:
            tool_obj.func = _make_wrapped_func(func)
        return tool_obj