Agentic Tool Calling Examples
Copy for LLM
Copy page as Markdown for LLMs
View as Markdown
Open this page as Markdown
Open in ChatGPT
Get insights from ChatGPT
Open in Claude
Get insights from Claude

Complete, ready-to-run examples demonstrating agentic workflows for common use cases.

Weather Report Workflow

Get weather information for multiple cities with automatic multi-step execution.

Python

import os
import requests

# Set your API key
API_KEY = os.environ['FREDDY_API_KEY']
ORG_ID = "your_org_id"

# Define tools
tools = [
 {
 "type": "function",
 "name": "get_cities",
 "description": "Get a list of 5 major cities for a specific geographic region",
 "parameters": {
 "type": "object",
 "properties": {
 "region": {
 "type": "string",
 "enum": ["asia", "europe", "americas", "africa", "oceania"],
 "description": "Geographic region to get cities from"
 }
 },
 "required": ["region"]
 }
 },
 {
 "type": "function",
 "name": "get_temperature",
 "description": "Get current temperature in Celsius for a specific city",
 "parameters": {
 "type": "object",
 "properties": {
 "city": {
 "type": "string",
 "description": "City name (e.g., 'Paris', 'Tokyo')"
 }
 },
 "required": ["city"]
 }
 }
]

# Single API call - backend handles everything!
response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={
 "X-API-Key": API_KEY,
 "Content-Type": "application/json"
 },
 json={
 "organizationId": ORG_ID,
 "model": "gpt-5",
 "inputs": [
 {
 "role": "user",
 "content": "Get a comprehensive weather report for major European cities"
 }
 ],
 "functions": tools,
 "parallelToolCalls": False,
 "maxToolCalls": 10
 }
)

result = response.json()

print("="*80)
print("WEATHER REPORT")
print("="*80)
print(result['response'])
print("\n" + "="*80)
print(f"Iterations: {result['metadata']['iterations']}")
print(f"Tool Calls: {result['metadata']['tool_calls']}")
print(f"Execution Time: {result['metadata']['execution_time_ms']}ms")

JavaScript/Node.js

const fetch = require('node-fetch');

const API_KEY = process.env.FREDDY_API_KEY;
const ORG_ID = 'your_org_id';

// Define tools
const tools = [
 {
 type: 'function',
 name: 'get_cities',
 description: 'Get a list of 5 major cities for a specific geographic region',
 parameters: {
 type: 'object',
 properties: {
 region: {
 type: 'string',
 enum: ['asia', 'europe', 'americas', 'africa', 'oceania'],
 description: 'Geographic region to get cities from'
 }
 },
 required: ['region']
 }
 },
 {
 type: 'function',
 name: 'get_temperature',
 description: 'Get current temperature in Celsius for a specific city',
 parameters: {
 type: 'object',
 properties: {
 city: {
 type: 'string',
 description: 'City name (e.g., "Paris", "Tokyo")'
 }
 },
 required: ['city']
 }
 }
];

async function getWeatherReport() {
 const response = await fetch('https://api.aitronos.com/v1/model/response', {
 method: 'POST',
 headers: {
 'X-API-Key': API_KEY,
 'Content-Type': 'application/json'
 },
 body: JSON.stringify({
 organizationId: ORG_ID,
 model: 'gpt-5',
 inputs: [
 {
 role: 'user',
 texts: [{
 text: 'Get a comprehensive weather report for major European cities'
 }]
 }
 ],
 functions: tools,
 parallelToolCalls: false,
 maxToolCalls: 10
 })
 });

 const result = await response.json();

 console.log('='.repeat(80));
 console.log('WEATHER REPORT');
 console.log('='.repeat(80));
 console.log(result.response[0].text); // Access text content
 console.log('\n' + '='.repeat(80));
 console.log(`Iterations: ${result.metadata.iterations}`);
 console.log(`Tool Calls: ${result.metadata.tool_calls}`);
 console.log(`Execution Time: ${result.metadata.execution_time_ms}ms`);
}

getWeatherReport();

Multi-Region Comparison

Compare weather across different regions using parallel execution (Claude).

Python

import requests
import os

response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "claude-sonnet-4-20250514", # Parallel execution
 "inputs": [
 {
 "role": "user",
 "content": "Compare the weather in Asia, Europe, and Americas. Get the temperature for all major cities and create a regional comparison report."
 }
 ],
 "functions": [
 {
 "type": "function",
 "name": "get_cities",
 "description": "Get 5 cities for a region",
 "parameters": {
 "type": "object",
 "properties": {
 "region": {
 "type": "string",
 "enum": ["asia", "europe", "americas"]
 }
 },
 "required": ["region"]
 }
 },
 {
 "type": "function",
 "name": "get_temperature",
 "description": "Get temperature for a city",
 "parameters": {
 "type": "object",
 "properties": {
 "city": {"type": "string"}
 },
 "required": ["city"]
 }
 }
 ],
 "parallelToolCalls": True, # Claude can handle parallel calls!
 "maxToolCalls": 20
 }
)

result = response.json()
print(result['response'])

# Claude efficiently handles:
# 1. Call get_cities for all 3 regions (parallel)
# 2. Call get_temperature for all 15 cities (parallel batches)
# 3. Generate comprehensive comparison

Research and Summarize

Combine web search with data processing.

Python

import requests
import os

response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "gpt-4o",
 "inputs": [
 {
 "role": "user",
 "content": "Research the latest AI developments in 2025 and create a summary with key trends"
 }
 ],
 "tools": [
 {
 "type": "webSearch", # Built-in tool!
 "searchContextSize": "high"
 }
 ],
 "maxToolCalls": 5
 }
)

result = response.json()
print(result['response'])

# Backend automatically:
# 1. Performs web search for "AI developments 2025"
# 2. Analyzes search results
# 3. May perform additional targeted searches
# 4. Synthesizes findings into summary

Data Analysis with Code Interpreter

Analyze data and create visualizations.

Python

import requests
import os

response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "gpt-4o",
 "inputs": [
 {
 "role": "user",
 "content": "Analyze the sales data in sales.csv and create a chart showing monthly revenue trends",
 "files": [{"fileId": "file_sales_csv_123"}]
 }
 ],
 "tools": [
 {
 "type": "codeInterpreter",
 "container": {
 "type": "auto",
 "fileIds": ["file_sales_csv_123"]
 }
 }
 ],
 "include": ["code_interpreter.outputs"] # Get generated files
 }
)

result = response.json()
print(result['response'])

# Access generated chart
for output in result.get('outputs', []):
 if output['type'] == 'image':
 print(f"Chart URL: {output['imageUrl']}")

# Backend automatically:
# 1. Loads sales.csv into Python environment
# 2. Analyzes data with pandas
# 3. Creates visualization with matplotlib
# 4. Generates insights report

Conditional Workflow

Execute different actions based on conditions.

Python

import requests
import os

def check_and_reorder():
 """Check inventory and reorder if stock is low"""

 response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "gpt-5",
 "inputs": [
 {
 "role": "user",
 "content": "Check inventory for product WIDGET-100. If stock is below 10 units, automatically place a reorder for 50 units."
 }
 ],
 "functions": [
 {
 "type": "function",
 "name": "check_inventory",
 "description": "Check current stock level for a product",
 "parameters": {
 "type": "object",
 "properties": {
 "product_id": {"type": "string"}
 },
 "required": ["product_id"]
 }
 },
 {
 "type": "function",
 "name": "place_reorder",
 "description": "Place a reorder for a product",
 "parameters": {
 "type": "object",
 "properties": {
 "product_id": {"type": "string"},
 "quantity": {"type": "integer"}
 },
 "required": ["product_id", "quantity"]
 }
 }
 ],
 "parallelToolCalls": False
 }
 )

 result = response.json()
 print(result['response'])

 # Model automatically:
 # 1. Calls check_inventory("WIDGET-100")
 # 2. Sees stock is 5 (below threshold)
 # 3. Calls place_reorder("WIDGET-100", 50)
 # 4. Confirms action taken

check_and_reorder()

Multi-Turn Conversation

Continue a workflow across multiple requests using threads.

Python

import requests
import os

API_KEY = os.environ['FREDDY_API_KEY']
ORG_ID = "your_org_id"
THREAD_ID = "thrd_persistent_123"

# First request
response1 = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": API_KEY},
 json={
 "organizationId": ORG_ID,
 "model": "gpt-5",
 "threadId": THREAD_ID, # Create thread
 "inputs": [
 {
 "role": "user",
 "content": "Get weather for European cities"
 }
 ],
 "functions": [
 {"type": "function", "name": "get_cities", "parameters": {...}},
 {"type": "function", "name": "get_temperature", "parameters": {...}}
 ]
 }
)

print("First Response:")
print(response1.json()['response'])

# Follow-up request - uses same thread!
response2 = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": API_KEY},
 json={
 "organizationId": ORG_ID,
 "model": "gpt-5",
 "threadId": THREAD_ID, # Same thread
 "inputs": [
 {
 "role": "user",
 "content": "Now compare those with Asian cities"
 }
 ],
 "functions": [
 {"type": "function", "name": "get_cities", "parameters": {...}},
 {"type": "function", "name": "get_temperature", "parameters": {...}}
 ]
 }
)

print("\nFollow-up Response:")
print(response2.json()['response'])

# Model remembers:
# - Previous conversation
# - European cities and temperatures
# - Can make comparisons

Error Handling

Gracefully handle tool failures.

Python

import requests
import os

response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "gpt-5",
 "inputs": [
 {
 "role": "user",
 "content": "Get weather data from our weather service"
 }
 ],
 "functions": [
 {
 "type": "function",
 "name": "get_weather_primary",
 "description": "Primary weather service (may be unavailable)",
 "parameters": {
 "type": "object",
 "properties": {
 "city": {"type": "string"}
 },
 "required": ["city"]
 }
 },
 {
 "type": "function",
 "name": "get_weather_backup",
 "description": "Backup weather service (always available)",
 "parameters": {
 "type": "object",
 "properties": {
 "city": {"type": "string"}
 },
 "required": ["city"]
 }
 }
 ]
 }
)

result = response.json()
print(result['response'])

# Model automatically:
# 1. Tries get_weather_primary()
# 2. Receives error response
# 3. Switches to get_weather_backup()
# 4. Returns results successfully
#
# No manual error handling needed!

Streaming with Agentic Workflows

See progress as the workflow executes (advanced).

Python

import requests
import os
import json

response = requests.post(
 "https://api.aitronos.com/v1/model/response",
 headers={"X-API-Key": os.environ['FREDDY_API_KEY']},
 json={
 "organizationId": "your_org_id",
 "model": "gpt-5",
 "stream": True, # Enable streaming
 "inputs": [
 {
 "role": "user",
 "content": "Get weather for European cities"
 }
 ],
 "functions": [
 {"type": "function", "name": "get_cities", "parameters": {...}},
 {"type": "function", "name": "get_temperature", "parameters": {...}}
 ]
 },
 stream=True
)

print("Streaming workflow progress:\n")

for line in response.iter_lines():
 if line:
 line = line.decode('utf-8')
 if line.startswith('data: '):
 data = json.loads(line[6:])

 # Tool call event
 if data.get('type') == 'tool_call.started':
 print(f" Calling {data['tool']['name']}...")

 # Tool result event
 elif data.get('type') == 'tool_call.completed':
 print(f" {data['tool']['name']} completed")

 # Text delta
 elif data.get('type') == 'response.text.delta':
 print(data['delta'], end='', flush=True)

 # Final completion
 elif data.get('type') == 'response.completed':
 print("\n\n Workflow completed!")

Best Practices

1. Set Appropriate Limits

{
 "maxToolCalls": 15, # Prevent runaway execution
 "parallelToolCalls": False # Sequential for dependent tasks
}

2. Use Clear Descriptions

{
 "name": "get_customer_orders",
 "description": "Retrieve all orders for a specific customer. Returns order ID, date, total, and status for each order. Use this when you need to check order history or find specific purchases."
}

3. Combine Built-In and Custom Tools

{
 "tools": [
 {"type": "webSearch"}, # Built-in: Fast server-side
 {
 "type": "function", # Custom: Your business logic
 "name": "save_to_crm",
 "parameters": {...}
 }
 ]
}

4. Enable Detailed Logging for Development

{
 "include": [
 "function_calls.logs",
 "request.logs",
 "usage.detailed"
 ]
}

5. Use Threads for Contextual Workflows

{
 "threadId": "thrd_user_session_123", # Maintain context
 "threadContextMode": "smart" # Intelligent history management
}

Common Patterns

Pattern: Data Enrichment

# Get basic data then enrich with additional calls
{
 "inputs": [{
 "role": "user",
 "content": "Get customer profile and enrich with order history and preferences"
 }],
 "functions": [
 "get_customer_profile", # Returns basic info
 "get_order_history", # Called with customer_id from profile
 "get_preferences" # Called with customer_id from profile
 ]
}

Pattern: Validation and Action

# Validate before taking action
{
 "inputs": [{
 "role": "user",
 "content": "Process refund for order #12345 if it's eligible"
 }],
 "functions": [
 "check_refund_eligibility", # Validation
 "process_refund" # Action (only if eligible)
 ]
}

Pattern: Search and Summarize

# Search multiple sources and synthesize
{
 "tools": [
 {"type": "webSearch"},
 {"type": "fileSearch", "vectorStoreIds": ["vs_docs"]}
 ],
 "inputs": [{
 "role": "user",
 "content": "Find information about X from web and our docs, then summarize"
 }]
}

Troubleshooting

Issue: Model Not Using Tools

Symptoms: Model responds without calling tools

Solutions:

{
 "toolChoice": "required", # Force tool usage
 # Or be more specific in the prompt:
 "text": "Use the get_weather tool to check the temperature in Paris"
}

Issue: Too Many Tool Calls

Symptoms: Workflow takes too long, excessive iterations

Solutions:

{
 "maxToolCalls": 5, # Lower limit
 # Or improve tool descriptions to be more specific
}

Issue: Incomplete Results

Symptoms: Response doesn't include expected information

Solutions:

{
 "parallelToolCalls": False, # Ensure sequential processing
 # Or add more explicit instructions:
 "text": "Make sure to check temperature for ALL cities before responding"
}

Next Steps

Function Calling Guide - Comprehensive reference
Agentic Workflows Guide - Deep dive
API Reference - Complete API documentation
System Tools - Built-in tool capabilities

Agentic Tool Calling ExamplesCopyCopy for LLMCopy page as Markdown for LLMsView as MarkdownOpen this page as MarkdownOpen in ChatGPTGet insights from ChatGPTOpen in ClaudeGet insights from Claude

Weather Report Workflow

Python

JavaScript/Node.js

Multi-Region Comparison

Python

Research and Summarize

Python

Data Analysis with Code Interpreter

Python

Conditional Workflow

Python

Multi-Turn Conversation

Python

Error Handling

Python

Streaming with Agentic Workflows

Python

Best Practices

1. Set Appropriate Limits

2. Use Clear Descriptions

3. Combine Built-In and Custom Tools

4. Enable Detailed Logging for Development

5. Use Threads for Contextual Workflows

Common Patterns

Pattern: Data Enrichment

Pattern: Validation and Action

Pattern: Search and Summarize

Troubleshooting

Issue: Model Not Using Tools

Issue: Too Many Tool Calls

Issue: Incomplete Results

Next Steps

Was this helpful?

Agentic Tool Calling Examples
Copy for LLM
Copy page as Markdown for LLMs
View as Markdown
Open this page as Markdown
Open in ChatGPT
Get insights from ChatGPT
Open in Claude
Get insights from Claude