	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Anthropic Claude API > Claude Parallel Tool Calls

Claude Parallel Tool Calls

Author: Venkata Sudhakar

Parallel tool calls allow ShopMax India's Claude integration to fetch multiple pieces of data simultaneously in a single API round-trip. Instead of sequentially checking product price, stock, and delivery estimate one after another, Claude can request all three tools at once, cutting response latency significantly for data-heavy queries.

Claude automatically issues parallel tool calls when it determines multiple tools can run independently. The response content will contain multiple tool_use blocks with different IDs. Your code collects all of them, executes the corresponding functions (potentially in parallel using asyncio or threads), and returns all tool_result blocks in one user message. Claude then synthesizes the combined results into a final answer.

The following example shows ShopMax India fetching price, stock, and delivery info for a product in parallel. Three tool_use blocks arrive together, all three functions run concurrently, and a single tool_result batch is sent back.

import anthropic
import json
import asyncio

client = anthropic.Anthropic()

def get_price(product_id):
    prices = {"PROD-TV65": 89999, "PROD-WM8": 34999, "PROD-AC15": 42500}
    return {"product_id": product_id, "price_rs": prices.get(product_id, 0)}

def get_stock(product_id):
    stock = {"PROD-TV65": 12, "PROD-WM8": 5, "PROD-AC15": 23}
    return {"product_id": product_id, "units": stock.get(product_id, 0), "city": "Hyderabad"}

def get_delivery(product_id, city):
    return {"product_id": product_id, "city": city, "days": 2, "courier": "ShopMax Express"}

tools = [
    {
        "name": "get_price",
        "description": "Get current price for a ShopMax India product",
        "input_schema": {"type": "object", "properties": {"product_id": {"type": "string"}}, "required": ["product_id"]}
    },
    {
        "name": "get_stock",
        "description": "Get available stock for a ShopMax India product",
        "input_schema": {"type": "object", "properties": {"product_id": {"type": "string"}}, "required": ["product_id"]}
    },
    {
        "name": "get_delivery",
        "description": "Get delivery estimate for a product to a city",
        "input_schema": {"type": "object", "properties": {"product_id": {"type": "string"}, "city": {"type": "string"}}, "required": ["product_id", "city"]}
    }
]

def run_tool(name, inputs):
    if name == "get_price":
        return get_price(inputs["product_id"])
    elif name == "get_stock":
        return get_stock(inputs["product_id"])
    elif name == "get_delivery":
        return get_delivery(inputs["product_id"], inputs["city"])
    return {"error": "unknown tool"}

messages = [{"role": "user", "content": "For Sony 65-inch TV (PROD-TV65), what is the price, stock, and delivery time to Chennai?"}]

response = client.messages.create(
    model="claude-haiku-4-5",
    max_tokens=512,
    tools=tools,
    messages=messages
)

tool_results = []
for block in response.content:
    if block.type == "tool_use":
        print("Calling tool:", block.name, block.input)
        result = run_tool(block.name, block.input)
        tool_results.append({"type": "tool_result", "tool_use_id": block.id, "content": json.dumps(result)})

print("Parallel tools called:", len(tool_results))

messages.append({"role": "assistant", "content": response.content})
messages.append({"role": "user", "content": tool_results})

final = client.messages.create(model="claude-haiku-4-5", max_tokens=512, tools=tools, messages=messages)
for block in final.content:
    if hasattr(block, "text"):
        print(block.text)

It gives the following output,

Calling tool: get_price {'product_id': 'PROD-TV65'}
Calling tool: get_stock {'product_id': 'PROD-TV65'}
Calling tool: get_delivery {'product_id': 'PROD-TV65', 'city': 'Chennai'}
Parallel tools called: 3

Here is the information for the Sony 65-inch TV (PROD-TV65):
- Price: Rs 89,999
- Stock: 12 units available (Hyderabad warehouse)
- Delivery to Chennai: 2 days via ShopMax Express

Parallel tool execution is most valuable when your tools hit external APIs or databases with independent latency. Wrap run_tool calls in asyncio.gather or a ThreadPoolExecutor to execute them concurrently rather than sequentially. Avoid parallel calls when tools have dependencies - for example, do not fetch order status in parallel with placing an order for the same product. Monitor token usage as parallel tool calls still count toward the context window size.

Send your comments, suggestions or queries regarding this site to [email protected].