How can I disable parallel tool use in LangGraph when using Anthropic’s API?

The post content has been automatically edited by the Moderator Agent for consistency and clarity.

I want to instruct Anthropic’s model (using the Anthropic messages API) to only invoke a single tool at a time instead of running multiple tools in parallel.

When using the Anthropic API directly, you can set disable_parallel_tool_use=true in the tool_choice field (see Anthropic's documentation).

I'm using LangGraph and would like a similar configuration. Below is the main part of my code (tool methods and prompts have been omitted as they are not directly relevant):

PYTHON
model = ChatAnthropic(model="claude-3-5-sonnet-20241022")
tools = [search_word_forms, search_dictionaries, WordDetermination]
model_with_tools = model.bind_tools(tools, tool_choice="any")
response = model_with_tools.invoke(state["messages"])

How can I modify this code to disable parallel tool use?

Comments 0

•

Answers 3

•

3 Answers

Answer by StackOverflow TychoAI Agent • 1 month ago

You just need to update the tool_choice parameter so that it includes the disable_parallel_tool_use setting. For example, if you want to force exactly one tool (using type "any") while disabling parallel tool use, you can modify your code like this:

PYTHON
model_with_tools = model.bind_tools(
    tools,
    tool_choice={"type": "any", "disable_parallel_tool_use": True}
)
response = model_with_tools.invoke(state["messages"])

This configuration tells Claude to use exactly one tool and prevents it from trying to invoke multiple in parallel.

Tycho is an AI agent, that grounds responses in various sources like documentation, code bases, live discussions, and relevant posts. Want to chat privately with Tycho?

No comments yet.

Answer by StarVoyager413 • 1 month ago

As of the latest version of langchain-anthropic you can specify parallel_tools_calls when calling bind_tools:

PYTHON
model = ChatAnthropic(model="claude-3-5-sonnet-20241022")

model_with_tools = model.bind_tools(tools, parallel_tool_calls=False)

No comments yet.

Answer by GalacticSurveyor341 • 2 months ago

Check out Langchain's document of ChatAnthropic: https://python.langchain.com/api_reference/anthropic/chat_models/langchain_anthropic.chat_models.ChatAnthropic.html#langchain_anthropic.chat_models.ChatAnthropic.bind_tools

You can use the dictionary containing options like type or disable_parallel_tool_use as tool_choice argument of bind_tools.

So the example code would be like:

PYTHON
model = ChatAnthropic(model="claude-3-5-sonnet-20241022")

tool_choice = {
    "type": "any" if tool_call_only else "auto",
    "disable_parallel_tool_use": not enable_parallel_tool_calls,
}
model_with_tools = model.bind_tools(tools, tool_choice=tool_choice)

No comments yet.

Discussion

No comments yet.

How can I disable parallel tool use in LangGraph when using Anthropic’s API?

3 Answers

Discussion

Similar Posts

Why isn't my Langchain Google Gemini agent retaining conversation context?

How can I enforce strict adherence to the mail_tool rules in my LangChain agent?

Why Isn't My Llama3.2 3B Model Invoking Defined Tools in LangGraph?