Tool calls

On /v1/chat/completions, send tools and, if necessary, tool_choice. The two fields have the OpenAI format, and we send them to the model.

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Current weather for a city.",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]

reply = client.chat.completions.create(
    model="claude-sonnet-5-5",
    messages=[{"role": "user", "content": "What is the weather in Nairobi?"}],
    tools=tools,
)

call = reply.choices[0].message.tool_calls[0]
print(call.function.name, call.function.arguments)

When the model asks for a tool, finish_reason is tool_calls and content is empty. This means that the model wants you to run the function.

Send the result back

Add the message of the model to the conversation. Then add a message with role: "tool" and the tool_call_id of the call:

messages = [
    {"role": "user", "content": "What is the weather in Nairobi?"},
    reply.choices[0].message,
    {
        "role": "tool",
        "tool_call_id": call.id,
        "content": '{"temp_c": 24, "conditions": "clear"}',
    },
]

final = client.chat.completions.create(model="claude-sonnet-5-5", messages=messages, tools=tools)
print(final.choices[0].message.content)

Models that accept tools

Look at supported_parameters in GET /v1/models, or at the table on Models. If you send tools to a model that does not show tools, you get a 400 error that names the parameter. We do not charge for that request.

Tool calls in a stream

Tool calls also come in a stream. delta.tool_calls has an index, and the arguments come in small parts. The last frame has finish_reason: "tool_calls". The SDKs join the parts for you. See Streams.

Messages and Responses

On /v1/messages and /v1/responses, use the tool format of Anthropic or of OpenAI Responses. We send your tools to the model as you wrote them.