Anthropic Tool Calling

New to Exa? Try the Coding Agent Quickstart to get started in under a minute.

Claude’s tool use allows models to call functions that you define in your code. The Exa SDKs ship ready-made web search and page reading tools for Anthropic, so you don’t have to hand-write the tool schema, parse tool_use blocks, or format Exa results yourself.

Get started

Install the SDKs

pip install anthropic exa_py

npm install @anthropic-ai/sdk exa-js

Set up your API keys

Set the EXA_API_KEY and ANTHROPIC_API_KEY environment variables. Visit the Anthropic console and the Exa dashboard to generate your API keys.

Get your Exa API key

Add the Exa tools to your tool loop

Pass the tools in the request’s tools list, then hand the assistant message to handle_tool_use. It executes every tool_use block in the message and returns the matching tool_result blocks, ready to send back in the next user message.web_search searches the web for pages the model hasn’t seen; get_contents reads pages it already has URLs for, whether from an earlier search or from the user. Register either or both.

import anthropic
from exa_py import Exa

exa = Exa()  # reads EXA_API_KEY from the environment
claude = anthropic.Anthropic()

messages = [{"role": "user", "content": "What's the latest on AI chips?"}]

response = claude.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=messages,
    tools=[exa.anthropic.web_search(), exa.anthropic.get_contents()],
)

messages.append({"role": "assistant", "content": response.content})
messages.append(
    {"role": "user", "content": exa.anthropic.handle_tool_use(response)}
)

response = claude.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=messages,
    tools=[exa.anthropic.web_search(), exa.anthropic.get_contents()],
)
print(response.content[0].text)

import Anthropic from "@anthropic-ai/sdk";
import Exa from "exa-js";

const exa = new Exa(); // reads EXA_API_KEY from the environment
const anthropic = new Anthropic();

const messages = [
  { role: "user", content: "What's the latest on AI chips?" },
];

let response = await anthropic.messages.create({
  model: "claude-sonnet-4-6",
  max_tokens: 1024,
  messages,
  tools: [exa.anthropic.webSearch(), exa.anthropic.getContents()],
});

messages.push({ role: "assistant", content: response.content });
messages.push({
  role: "user",
  content: await exa.anthropic.handleToolUse(response),
});

response = await anthropic.messages.create({
  model: "claude-sonnet-4-6",
  max_tokens: 1024,
  messages,
  tools: [exa.anthropic.webSearch(), exa.anthropic.getContents()],
});
console.log(response.content[0].text);

This is one round for brevity. A real agent keeps tools on every request and repeats the handler step until the model replies without tool_use blocks — that’s how a search result turns into a follow-up page read.Calling the factories with no arguments gives Exa’s recommended defaults: type="auto" with contents={"highlights": True} for search, and page text capped at 10,000 characters for contents.

Configuring the tools

Keyword arguments are regular Exa options, passed through when the tool runs — search options to exa.search(), contents options to exa.get_contents():

tools = [
    exa.anthropic.web_search(category="news", contents={"text": True}),
    exa.anthropic.get_contents(summary=True, livecrawl="preferred"),
]

const tools = [
  exa.anthropic.webSearch({ category: "news", contents: { text: true } }),
  exa.anthropic.getContents({ summary: true, livecrawl: "preferred" }),
];

The model picks the search query and the urls to read; everything else is bound when you create the tool, so it can’t change what gets crawled or extracted. name (defaulting to "web_search" and "get_contents") and description instead override the tool definition the model sees. Anthropic requires tool names to be unique, so a custom name lets the Exa tool run alongside Anthropic’s built-in web_search_20250305 server tool, which reserves the web_search name:

response = claude.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=messages,
    tools=[
        exa.anthropic.web_search(name="exa_web_search"),
        {"type": "web_search_20250305", "name": "web_search", "max_uses": 5},
    ],
)

const response = await anthropic.messages.create({
  model: "claude-sonnet-4-6",
  max_tokens: 1024,
  messages,
  tools: [
    exa.anthropic.webSearch({ name: "exa_web_search" }),
    { type: "web_search_20250305", name: "web_search", max_uses: 5 },
  ],
});

Mixing in your own tools

handle_tool_use answers every tool_use block in the message: a block naming a tool it can’t resolve gets an Error: unknown tool "<name>" result instead of being dropped, so the follow-up request never omits a required tool result. If you run your own tools alongside Exa’s, replace those error results with your own before the next request.

Writing the loop by hand

If you’d rather own the tool schema and execution yourself, define the tool and process the tool_use blocks manually. exa.tools.web_search() and exa.tools.get_contents() give you the same provider-neutral tool specs (with a run method) for hand-rolled loops, or you can write everything from scratch:

Python

TOOLS = [
    {
        "name": "exa_search",
        "description": "Perform a search query on the web, and retrieve the most relevant URLs/web data.",
        "input_schema": {
            "type": "object",
            "properties": {
                "query": {
                    "type": "string",
                    "description": "The search query to perform.",
                },
            },
            "required": ["query"],
        },
    }
]

def exa_search(query: str):
    return exa.search(query=query, type="auto", contents={"highlights": True})

def process_tool_use(response):
    results = []
    for block in response.content:
        if block.type == "tool_use" and block.name == "exa_search":
            results.append(
                {
                    "type": "tool_result",
                    "tool_use_id": block.id,
                    "content": str(exa_search(**block.input)),
                }
            )
    return results

See the Python SDK specification and TypeScript SDK specification for the full search and contents options.

API

Data

MCP

SDKs

Integrations

Agentic payments

Admin

Additional Resources

Anthropic Tool Calling

Get started

Get your Exa API key

Configuring the tools

Mixing in your own tools

Writing the loop by hand

​Get started

Get your Exa API key

​Configuring the tools

​Mixing in your own tools

​Writing the loop by hand

Get started

Configuring the tools

Mixing in your own tools

Writing the loop by hand