Agent State

This example shows how to create an agent that maintains state across interactions. It demonstrates a simple counter mechanism, but this pattern can be extended to more complex state management like maintaining conversation context, user preferences, or tracking multi-step processes.

Example prompts to try:

• “Increment the counter 3 times and tell me the final count”
• “What’s our current count? Add 2 more to it”
• “Let’s increment the counter 5 times, but tell me each step”
• “Add 4 to our count and remind me where we started”
• “Increase the counter twice and summarize our journey”

Code

agent_state.py

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.run import RunContext


def increment_counter(run_context: RunContext) -> str:
    """Increment the counter in session state."""
    # Initialize counter if it doesn't exist
    if not run_context.session_state:
        run_context.session_state = {}

    if "count" not in run_context.session_state:
        run_context.session_state["count"] = 0

    # Increment the counter
    run_context.session_state["count"] += 1

    return f"Counter incremented! Current count: {run_context.session_state['count']}"


def get_counter(run_context: RunContext) -> str:
    """Get the current counter value."""
    if not run_context.session_state:
        run_context.session_state = {}

    count = run_context.session_state.get("count", 0)
    return f"Current count: {count}"


# Create an Agent that maintains state
agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    # Initialize the session state with a counter starting at 0
    session_state={"count": 0},
    tools=[increment_counter, get_counter],
    # Use variables from the session state in the instructions
    instructions="You can increment and check a counter. Current count is: {count}",
    # Important: Resolve the state in the messages so the agent can see state changes
    resolve_in_context=True,
    markdown=True,
)

# Test the counter functionality
print("Testing counter functionality...")
agent.print_response(
    "Let's increment the counter 3 times and observe the state changes!", stream=True
)
print(f"Final session state: {agent.get_session_state()}")

Usage

1. Create a virtual environment

   Open the Terminal and create a python virtual environment.

   Mac

   python3 -m venv .venv
   source .venv/bin/activate
   
2. Install libraries

   pip install openai agno
   
3. Run the agent

   python agent_state.py