Implementation:Langchain ai Langchain ChatAnthropicTools

Knowledge Sources	Langchain_ai_Langchain
Domains	Tool Calling, Anthropic, XML Parsing
Last Updated	2026-02-11 00:00 GMT

Overview

The experimental module provides XML-based tool-calling support for Anthropic chat models, generating system prompts with tool descriptions and parsing XML tool-call responses.

Description

This module, located in langchain_anthropic.experimental, implements an experimental mechanism for tool calling with Anthropic models using an XML-based format. It includes functions to generate system messages that describe available tools in XML, parse XML responses back into structured tool call dictionaries, and handle type coercion for tool parameters. The module uses XML templates (SYSTEM_PROMPT_FORMAT, TOOL_FORMAT, TOOL_PARAMETER_FORMAT) to format tool definitions into a system prompt that instructs the model on how to invoke tools.

Usage

Import the functions from this module when integrating Anthropic models with tool-calling capabilities using the XML-based invocation format. This is considered experimental functionality.

Code Reference

Source Location

• Repository: Langchain_ai_Langchain
• File: libs/partners/anthropic/langchain_anthropic/experimental.py
• Lines: 1-141

Signature

def get_system_message(tools: list[dict]) -> str: ...

def _get_type(parameter: dict[str, Any]) -> str: ...

def _xml_to_dict(t: Any) -> str | dict[str, Any]: ...

def _xml_to_function_call(invoke: Any, tools: list[dict]) -> dict[str, Any]: ...

def _xml_to_tool_calls(elem: Any, tools: list[dict]) -> list[dict[str, Any]]: ...

Import

from langchain_anthropic.experimental import get_system_message, _xml_to_tool_calls

I/O Contract

Inputs (get_system_message)

Name	Type	Required	Description
tools	list[dict]	Yes	List of tool definitions, each containing name, description, and parameters (with properties sub-dict).

Inputs (_xml_to_tool_calls)

Name	Type	Required	Description
elem	Any	Yes	An XML element containing invoke child elements.
tools	list[dict]	Yes	The list of tool definitions to match against for type coercion.

Outputs

Name	Type	Description
get_system_message return	str	A formatted system message string describing available tools in XML format.
_xml_to_tool_calls return	list[dict[str, Any]]	List of tool call dictionaries, each with function (containing name and arguments) and type keys.

XML Templates

The module defines three XML template constants:

• SYSTEM_PROMPT_FORMAT -- The outer template that wraps all tool descriptions and instructs the model on the invocation format.
• TOOL_FORMAT -- Template for each individual tool description including name, description, and parameters.
• TOOL_PARAMETER_FORMAT -- Template for each parameter within a tool, including name, type, and description.

Usage Examples

Basic Usage

from langchain_anthropic.experimental import get_system_message

tools = [
    {
        "name": "get_weather",
        "description": "Get the current weather for a location",
        "parameters": {
            "properties": {
                "location": {
                    "type": "string",
                    "description": "The city and state"
                }
            }
        }
    }
]

system_msg = get_system_message(tools)
# Returns a formatted XML system prompt describing the available tools

Related Pages

• Environment:Langchain_ai_Langchain_Anthropic_API_Credentials
• Part of the langchain-anthropic partner package