Build Your First Agent

Build and run your first agent in 20 lines of code.

Let’s build an agent that can read/write local files, has long and short term memory, and runs as a production API you can build a product on top of.

1. Create the Agent

Save the following code as workbench.py:

workbench.py

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.os import AgentOS
from agno.tools.workspace import Workspace

workbench = Agent(
    name="Workbench",
    model="openai:gpt-5.4",
    tools=[Workspace(".",
        allowed=["read", "list", "search"],
        confirm=["write", "edit", "delete", "shell"],
    )],  # read/write/edit/shell in this directory
    enable_agentic_memory=True,  # remembers across sessions
    add_history_to_context=True,  # include past runs
    num_history_runs=3,  # last 3 conversations
)

# Serve via AgentOS
agent_os = AgentOS(agents=[workbench], tracing=True, db=SqliteDb(db_file="agno.db"))
app = agent_os.get_app()

Workspace(".") scopes the agent to the current directory. read, list, and search run without approval. write, edit, delete, and shell require human approval.

2. Run Your AgentOS

1. Set up your virtual environment

   Mac

   uv venv --python 3.12
   source .venv/bin/activate
   

   uv venv --python 3.12
   .venv\Scripts\activate
   
2. Install dependencies

   uv pip install -U 'agno[os]' openai
   
3. Export your OpenAI API key

   Don’t have one? Get a key from platform.openai.com.

   Mac

   export OPENAI_API_KEY=sk-***
   

   setx OPENAI_API_KEY sk-***
   
4. Run your AgentOS

   fastapi dev workbench.py
   

   Your AgentOS is now running at:

   http://localhost:8000

   API documentation is automatically available at:

   http://localhost:8000/docs

Now you have:

• A stateful agent served as a production API.
• Tracing and session monitoring enabled out of the box.
• Per-session storage and isolation, with JWT-based RBAC available for multi-user isolation.

No third-party services required.

3. Chat with your Agent

The AgentOS UI connects your browser to your runtime.

Use it to test, monitor, and manage your agents in real time.

1. Open os.agno.com and sign in.
2. Click “Connect OS”
3. Select “Local”, enter your endpoint URL (default: http://localhost:8000), name it “Local AgentOS”, and click “Connect”.

Click on Chat, and ask:

List the files here and tell me what kind of project this is.

The agent reads your workspace and answers grounded in what it actually finds.

Try a follow-up like “summarize the README” or “create a NOTES.md with three bullet takeaways”. The second one will pause for your approval before the file is written, since write_file is a confirm-required tool by default.

Click Sessions or Traces in the sidebar to inspect stored conversations.

All session data is stored in your local database, no data leaves your system.

Next

Build an agent platform

The full template: clone, ship agents with Claude Code, deploy to Railway.

Start from a production template

Scout, Dash, Coda. Pick the closest to what you’re building.

Understand the runtime

Why AgentOS exists and the work it handles for you.

Use Agno with your coding agent

Wire up agno documentation with your favorite coding agent.