Responses API Beta Tool Calling - Function Calling Integration

Beta APIThis API is in beta stage and may have breaking changes.

The Responses API Beta supports comprehensive tool calling capabilities, allowing models to call functions, execute tools in parallel, and handle complex multi-step workflows.

Basic Tool Definition

Define tools using the OpenAI function calling format:

const weatherTool = {
  type: 'function' as const,
  name: 'get_weather',
  description: 'Get the current weather in a location',
  strict: null,
  parameters: {
    type: 'object',
    properties: {
      location: {
        type: 'string',
        description: 'The city and state, e.g. San Francisco, CA',
      },
      unit: {
        type: 'string',
        enum: ['celsius', 'fahrenheit'],
      },
    },
    required: ['location'],
  },
};

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'What is the weather in San Francisco?',
          },
        ],
      },
    ],
    tools: [weatherTool],
    tool_choice: 'auto',
    max_output_tokens: 9000,
  }),
});

const result = await response.json();
console.log(result);

Tool Choice Options

Control when and how tools are called:

Tool Choice	Description
`auto`	Model decides whether to call tools
`none`	Model will not call any tools
`{type: 'function', name: 'tool_name'}`	Force specific tool call

Force Specific Tool

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'Hello, how are you?',
          },
        ],
      },
    ],
    tools: [weatherTool],
    tool_choice: { type: 'function', name: 'get_weather' },
    max_output_tokens: 9000,
  }),
});

Disable Tool Calling

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'What is the weather in Paris?',
          },
        ],
      },
    ],
    tools: [weatherTool],
    tool_choice: 'none',
    max_output_tokens: 9000,
  }),
});

Multiple Tools

Define multiple tools for complex workflows:

const calculatorTool = {
  type: 'function' as const,
  name: 'calculate',
  description: 'Perform mathematical calculations',
  strict: null,
  parameters: {
    type: 'object',
    properties: {
      expression: {
        type: 'string',
        description: 'The mathematical expression to evaluate',
      },
    },
    required: ['expression'],
  },
};

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'What is 25 * 4?',
          },
        ],
      },
    ],
    tools: [weatherTool, calculatorTool],
    tool_choice: 'auto',
    max_output_tokens: 9000,
  }),
});

Parallel Tool Calls

The API supports parallel execution of multiple tools:

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'Calculate 10*5 and also tell me the weather in Miami',
          },
        ],
      },
    ],
    tools: [weatherTool, calculatorTool],
    tool_choice: 'auto',
    max_output_tokens: 9000,
  }),
});

const result = await response.json();
console.log(result);

Tool Call Response

When tools are called, the response includes function call information:

{
  "id": "resp_1234567890",
  "object": "response",
  "created_at": 1234567890,
  "model": "openai/o4-mini",
  "output": [
    {
      "type": "function_call",
      "id": "fc_abc123",
      "call_id": "call_xyz789",
      "name": "get_weather",
      "arguments": "{\"location\":\"San Francisco, CA\"}"
    }
  ],
  "usage": {
    "input_tokens": 45,
    "output_tokens": 25,
    "total_tokens": 70
  },
  "status": "completed"
}

Tool Responses in Conversation

Include tool responses in follow-up requests:

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'What is the weather in Boston?',
          },
        ],
      },
      {
        type: 'function_call',
        id: 'fc_1',
        call_id: 'call_123',
        name: 'get_weather',
        arguments: JSON.stringify({ location: 'Boston, MA' }),
      },
      {
        type: 'function_call_output',
        id: 'fc_output_1',
        call_id: 'call_123',
        output: JSON.stringify({ temperature: '72°F', condition: 'Sunny' }),
      },
      {
        type: 'message',
        role: 'assistant',
        id: 'msg_abc123',
        status: 'completed',
        content: [
          {
            type: 'output_text',
            text: 'The weather in Boston is currently 72°F and sunny. This looks like perfect weather for a picnic!',
            annotations: []
          }
        ]
      },
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'Is that good weather for a picnic?',
          },
        ],
      },
    ],
    max_output_tokens: 9000,
  }),
});

Optional FieldThe id field is optional for function_call_output objects. Only type, call_id, and output are required — call_id is what pairs the output with its originating function_call. The examples above include id for completeness, but you can safely omit it.

Multimodal tool outputs

function_call_output.output accepts either a string or an array of input content parts (input_text, input_image, input_file) — the same shape as user-message content. Use the array form to return images or files from a tool; the non-text parts are only forwarded to supported multimodal models.

{
  "type": "function_call_output",
  "call_id": "call_123",
  "output": [
    {
      "type": "input_text",
      "text": "{\"results\":[{\"title\":\"Golden Gate Bridge\"}]}"
    },
    {
      "type": "input_image",
      "image_url": "https://example.com/image.jpg"
    }
  ]
}

Streaming Tool Calls

Monitor tool calls in real-time with streaming:

const response = await fetch('https://openrouter.ai/api/v1/responses', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_OPENROUTER_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/o4-mini',
    input: [
      {
        type: 'message',
        role: 'user',
        content: [
          {
            type: 'input_text',
            text: 'What is the weather like in Tokyo, Japan? Please check the weather.',
          },
        ],
      },
    ],
    tools: [weatherTool],
    tool_choice: 'auto',
    stream: true,
    max_output_tokens: 9000,
  }),
});

const reader = response.body?.getReader();
const decoder = new TextDecoder();

while (true) {
  const { done, value } = await reader.read();
  if (done) break;

  const chunk = decoder.decode(value);
  const lines = chunk.split('\n');

  for (const line of lines) {
    if (line.startsWith('data: ')) {
      const data = line.slice(6);
      if (data === '[DONE]') return;

      try {
        const parsed = JSON.parse(data);
        if (parsed.type === 'response.output_item.added' &&
            parsed.item?.type === 'function_call') {
          console.log('Function call:', parsed.item.name);
        }
        if (parsed.type === 'response.function_call_arguments.done') {
          console.log('Arguments:', parsed.arguments);
        }
      } catch (e) {
        // Skip invalid JSON
      }
    }
  }
}

Tool Validation

Ensure tool calls have proper structure:

{
  "type": "function_call",
  "id": "fc_abc123",
  "call_id": "call_xyz789",
  "name": "get_weather",
  "arguments": "{\"location\":\"Seattle, WA\"}"
}

Required fields:

type: Always “function_call”
id: Unique identifier for the function call object
name: Function name matching tool definition
arguments: Valid JSON string with function parameters
call_id: Unique identifier for the call

Best Practices

Clear descriptions: Provide detailed function descriptions and parameter explanations
Proper schemas: Use valid JSON Schema for parameters
Error handling: Handle cases where tools might not be called
Parallel execution: Design tools to work independently when possible
Conversation flow: Include tool responses in follow-up requests for context

Next Steps

Learn about Web Search integration
Explore Reasoning with tools
Review Basic Usage fundamentals

​Basic Tool Definition

​Tool Choice Options

​Force Specific Tool

​Disable Tool Calling

​Multiple Tools

​Parallel Tool Calls

​Tool Call Response

​Tool Responses in Conversation

​Multimodal tool outputs

​Streaming Tool Calls

​Tool Validation

​Best Practices

​Next Steps

Basic Tool Definition

Tool Choice Options

Force Specific Tool

Disable Tool Calling

Multiple Tools

Parallel Tool Calls

Tool Call Response

Tool Responses in Conversation

Multimodal tool outputs

Streaming Tool Calls

Tool Validation

Best Practices

Next Steps