Tool Calling

Tool calling lets the model decide to invoke one of your functions to fetch data or take action — for example, looking up the weather, querying a database, or sending an email. The model doesn’t execute the tool itself; it returns the function name and arguments, your code runs the function, and you send the result back in a follow-up message.

This guide covers function tools that your application executes. For tools executed automatically by Mixlayer, see Server-Side Tools.

Mixlayer supports tool calling through both Chat Completions and Responses. The model behavior is the same, but the request and follow-up item formats differ.

The loop

1. Define your tools in the tools array on the request.
2. Send the request. If the model decides to invoke a tool, Chat Completions returns finish_reason: "tool_calls" with an assistant tool_calls array. Responses returns a function_call output item.
3. Execute the tool in your code.
4. Send a follow-up request with the tool result. Chat Completions uses a tool role message. Responses uses a function_call_output input item.
5. The model returns its final answer, conditioned on the tool result.

Defining tools

The parameters field is a JSON Schema describing the tool’s arguments.

1
{
2
  "type": "function",
3
  "function": {
4
    "name": "get_weather",
5
    "description": "Get the current weather for a city.",
6
    "parameters": {
7
      "type": "object",
8
      "properties": {
9
        "city": {
10
          "type": "string",
11
          "description": "City name, e.g. 'San Francisco'"
12
        }
13
      },
14
      "required": ["city"]
15
    },
16
    "strict": true
17
  }
18
}

Set strict: true if you want the model’s arguments to be guaranteed to match the schema.

Tool names must be 1-64 characters and contain only ASCII letters, numbers, _, or -.

Sending Tool Results

1
{
2
  "messages": [
3
    { "role": "user", "content": "What is the weather in Paris?" },
4
    {
5
      "role": "assistant",
6
      "tool_calls": [{
7
        "id": "call_abc123",
8
        "type": "function",
9
        "function": {
10
          "name": "get_weather",
11
          "arguments": "{\"city\":\"Paris\"}"
12
        }
13
      }]
14
    },
15
    {
16
      "role": "tool",
17
      "tool_call_id": "call_abc123",
18
      "content": "{\"temp_c\":18,\"condition\":\"cloudy\"}"
19
    }
20
  ]
21
}

Worked example: Chat Completions weather lookup

The full loop, end to end:

$
# Step 1: Send the request with tools defined.
$
curl https://models.mixlayer.ai/v1/chat/completions \
>
  -H "Authorization: Bearer $MIXLAYER_API_KEY" \
>
  -H "Content-Type: application/json" \
>
  -d '{
>
    "model": "qwen/qwen3.6-27b",
>
    "messages": [
>
      {"role": "user", "content": "What is the weather in Paris?"}
>
    ],
>
    "tools": [{
>
      "type": "function",
>
      "function": {
>
        "name": "get_weather",
>
        "description": "Get the current weather for a city.",
>
        "parameters": {
>
          "type": "object",
>
          "properties": {"city": {"type": "string"}},
>
          "required": ["city"]
>
        }
>
      }
>
    }]
>
  }'
$
$
# Response includes:
$
#   "tool_calls": [{
>
#     "id": "call_abc123",
>
#     "type": "function",
>
#     "function": {"name": "get_weather", "arguments": "{\"city\":\"Paris\"}"}
>
#   }]
$
# and "finish_reason": "tool_calls"
$
$
# Step 2: Run the tool yourself, then send a follow-up with the result.
$
curl https://models.mixlayer.ai/v1/chat/completions \
>
  -H "Authorization: Bearer $MIXLAYER_API_KEY" \
>
  -H "Content-Type: application/json" \
>
  -d '{
>
    "model": "qwen/qwen3.6-27b",
>
    "messages": [
>
      {"role": "user", "content": "What is the weather in Paris?"},
>
      {"role": "assistant", "tool_calls": [{
>
        "id": "call_abc123",
>
        "type": "function",
>
        "function": {"name": "get_weather", "arguments": "{\"city\":\"Paris\"}"}
>
      }]},
>
      {"role": "tool", "tool_call_id": "call_abc123", "content": "{\"temp_c\": 18, \"condition\": \"cloudy\"}"}
>
    ]
>
  }'

Streaming with tool calls

When stream: true and the model invokes a tool, tool_calls arrive in deltas across multiple chunks. Each chunk’s delta.tool_calls[].function.arguments is a partial JSON string fragment — accumulate them per tool_calls[].index until the final chunk arrives with finish_reason: "tool_calls".

data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_abc","function":{"name":"get_weather","arguments":""}}]}}]}
data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"city\":"}}]}}]}
data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"Paris\"}"}}]}}]}
data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}]}

Notes

Tool calling is supported on the Qwen 3.5 family. See Models for the up-to-date list of capable models.