AI Agents & Agentic Workflows

Agent vs. chatbot — the key difference:

Chatbot (single turn): User: "What's the weather in Tokyo?" LLM: "I don't have access to real-time weather data." Agent (multi-step): User: "What's the weather in Tokyo?" Agent thinks: "I need to check weather data. I have a weather API tool." Agent calls: weather_api(city="Tokyo") Agent observes: {"temp": 18, "condition": "partly cloudy", "humidity": 65} Agent responds: "It's 18°C and partly cloudy in Tokyo with 65% humidity."

How tool use works under the hood:

# 1. Define tools as structured schemas tools = [ { "name": "search_database", "description": "Search the product database by query", "parameters": { "type": "object", "properties": { "query": {"type": "string", "description": "Search query"}, "limit": {"type": "integer", "description": "Max results"} }, "required": ["query"] } }, { "name": "send_email", "description": "Send an email to a customer", "parameters": { "type": "object", "properties": { "to": {"type": "string"}, "subject": {"type": "string"}, "body": {"type": "string"} }, "required": ["to", "subject", "body"] } } ] # 2. LLM sees the tools and decides to call one response = llm.generate( messages=[{"role": "user", "content": "Find laptops under $1000"}], tools=tools ) # 3. LLM returns a tool call (not text!) # response.tool_calls = [ # {"name": "search_database", "arguments": {"query": "laptops under $1000", "limit": 5}} # ] # 4. Your code executes the tool and feeds the result back result = search_database(query="laptops under $1000", limit=5) # Continue the conversation with the tool result...

The ReAct loop:

Task: "Book a meeting with Sarah next Tuesday at 2pm and send her a confirmation." Step 1 - REASON: I need to check Sarah's calendar availability for next Tuesday at 2pm. Step 1 - ACT: check_calendar(user="sarah", date="2026-03-24", time="14:00") Step 1 - OBSERVE: {"available": true, "conflicts": []} Step 2 - REASON: Sarah is available. Now I need to create the calendar event. Step 2 - ACT: create_event(title="Meeting", attendees=["sarah"], date="2026-03-24", time="14:00") Step 2 - OBSERVE: {"event_id": "evt_123", "status": "confirmed"} Step 3 - REASON: Event created. Now I need to send a confirmation email. Step 3 - ACT: send_email(to="sarah@company.com", subject="Meeting Confirmed - Tue Mar 24 at 2pm", body="...") Step 3 - OBSERVE: {"status": "sent"} Step 4 - REASON: All tasks complete. I can respond to the user. FINAL: "Done! I've booked a meeting with Sarah for next Tuesday at 2pm and sent her a confirmation email."

LangGraph agent example (conceptual):

from langgraph.graph import StateGraph, MessagesState, START, END from langchain_anthropic import ChatAnthropic # Define the LLM with tools model = ChatAnthropic(model="claude-sonnet-4-20250514").bind_tools(tools) # Define the agent's reasoning node def agent_node(state: MessagesState): response = model.invoke(state["messages"]) return {"messages": [response]} # Define the tool execution node def tool_node(state: MessagesState): # Execute the tool calls from the last message results = execute_tools(state["messages"][-1].tool_calls) return {"messages": results} # Build the graph graph = StateGraph(MessagesState) graph.add_node("agent", agent_node) graph.add_node("tools", tool_node) # Define the flow: start → agent → tools → agent → ... → end graph.add_edge(START, "agent") graph.add_conditional_edges("agent", should_continue, {"tools": "tools", "end": END}) graph.add_edge("tools", "agent") # Compile and run app = graph.compile() result = app.invoke({"messages": [("user", "What's the weather in Tokyo?")]})

CrewAI multi-agent example (conceptual):

from crewai import Agent, Task, Crew # Define agents with roles, goals, and backstories researcher = Agent( role="Senior Research Analyst", goal="Find comprehensive, accurate information on the given topic", backstory="You are an experienced research analyst with a keen eye " "for reliable sources and data-driven insights.", tools=[web_search, arxiv_search], ) writer = Agent( role="Technical Writer", goal="Create clear, engaging content from research findings", backstory="You are a skilled technical writer who excels at making " "complex topics accessible to a broad audience.", tools=[text_editor], ) # Define tasks research_task = Task( description="Research the latest developments in AI agents for 2026", agent=researcher, expected_output="A detailed research brief with key findings and sources" ) writing_task = Task( description="Write a blog post based on the research brief", agent=writer, expected_output="A polished 1500-word blog post", context=[research_task] # This task depends on the research task ) # Assemble the crew and run crew = Crew(agents=[researcher, writer], tasks=[research_task, writing_task]) result = crew.kickoff()

Anthropic Agent SDK example (conceptual, Python):

import anthropic from anthropic.types import ToolUseBlock client = anthropic.Anthropic() # Define tools tools = [ { "name": "get_weather", "description": "Get current weather for a city", "input_schema": { "type": "object", "properties": { "city": {"type": "string", "description": "City name"} }, "required": ["city"] } } ] # Agent loop: reason → act → observe → repeat messages = [{"role": "user", "content": "What's the weather in Tokyo and London?"}] while True: response = client.messages.create( model="claude-sonnet-4-20250514", max_tokens=1024, tools=tools, messages=messages, ) # If the model wants to use tools, execute them if response.stop_reason == "tool_use": tool_results = [] for block in response.content: if isinstance(block, ToolUseBlock): result = execute_tool(block.name, block.input) tool_results.append({ "type": "tool_result", "tool_use_id": block.id, "content": str(result), }) messages.append({"role": "assistant", "content": response.content}) messages.append({"role": "user", "content": tool_results}) else: # Model is done — print final response print(response.content[0].text) break

Example: customer support multi-agent system

┌─────────────┐ User message ───→ │ Router │ │ Agent │ └──────┬──────┘ │ ┌────────────┼────────────┐ ▼ ▼ ▼ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ Billing │ │ Technical│ │ Sales │ │ Agent │ │ Agent │ │ Agent │ │ │ │ │ │ │ │ Tools: │ │ Tools: │ │ Tools: │ │ - orders │ │ - logs │ │ - CRM │ │ - refund │ │ - docs │ │ - pricing│ │ - invoice│ │ - tickets│ │ - demos │ └─────┬────┘ └─────┬────┘ └─────┬────┘ │ │ │ └────────────┼────────────┘ ▼ ┌─────────────┐ │ Quality │ │ Checker │ │ Agent │ └──────┬──────┘ │ ▼ Response to user