Your first agent

This page builds a working agent backed by a real language model. You will write the graph, run it in Python, and see a response.

What you need

• AgentFlow installed: pip install 10xscale-agentflow
• A language model API key (this example uses Google Gemini)

Set your API key:

export GOOGLE_API_KEY=your-api-key

The graph

Create first_agent.py:

from agentflow.core.graph import Agent, StateGraph, ToolNode
from agentflow.core.state import AgentState
from agentflow.utils import END

# Create the agent node backed by a language model
agent = Agent(
    model="google/gemini-2.5-flash",
    system_prompt=[
        {
            "role": "system",
            "content": "You are a helpful assistant. Answer questions clearly and concisely.",
        }
    ],
)

# Build the graph
graph = StateGraph(AgentState)
graph.add_node("assistant", agent)
graph.set_entry_point("assistant")
graph.add_edge("assistant", END)

app = graph.compile()

Run it

Add the invocation code at the bottom of first_agent.py:

from agentflow.core.state import Message

result = app.invoke(
    {"messages": [Message.text_message("What is the capital of France?")]},
    config={"thread_id": "beginner-demo-1"},
)

print(result["messages"][-1].text())

Run the file:

python first_agent.py

Expected output (the exact words will vary):

The capital of France is Paris.

What happened

1. app.invoke adds your message to AgentState.context.
2. The graph routes to the assistant node.
3. Agent sends the conversation to the language model.
4. The model returns a reply, which becomes an assistant Message.
5. The graph reaches END and returns the updated state.

The thread_id in config groups this conversation. Every call with the same thread_id will eventually share history once you add a checkpointer.

Key imports

from agentflow.core.graph import Agent, StateGraph
from agentflow.core.state import AgentState, Message
from agentflow.utils import END

What you learned

• Agent wraps a language model as a graph node.
• StateGraph wires the node into a runnable app.
• app.invoke runs the graph and returns updated state.
• A thread_id in config identifies the conversation.

Next step

Give the agent something it can do — add a tool.