Introduction

Build agents, teams, and workflows in pure python.

The Agno SDK gives you three primitives: Agent, Team, Workflow, and a host of capabilities to turn them into production systems.

Agent

from agno.agent import Agent
from agno.tools.yfinance import YFinanceTools

agent = Agent(
    name="Finance Agent",
    model="openai:gpt-5.4",
    tools=[YFinanceTools()],
    instructions="Fetch market data and produce a one-line take.",
)

agent.print_response("What's NVDA trading at today?")

from agno.agent import Agent
from agno.team import Team
from agno.tools.yfinance import YFinanceTools

bull = Agent(
    name="Bull",
    model="openai:gpt-5.4",
    role="Make the case FOR investing.",
    tools=[YFinanceTools()],
)
bear = Agent(
    name="Bear",
    model="openai:gpt-5.4",
    role="Make the case AGAINST investing.",
    tools=[YFinanceTools()],
)

team = Team(
    name="Investment Committee",
    members=[bull, bear],
    instructions="Hear both sides, then synthesize a balanced recommendation.",
)

team.print_response("Should I invest in NVIDIA?")

from agno.agent import Agent
from agno.team import Team
from agno.tools.yfinance import YFinanceTools
from agno.workflow import Step, Workflow

researcher = Agent(
    model="openai:gpt-5.4",
    tools=[YFinanceTools()],
    instructions="Gather raw market data.",
)

bull = Agent(model="openai:gpt-5.4", role="Make the case FOR investing.")
bear = Agent(model="openai:gpt-5.4", role="Make the case AGAINST investing.")
committee = Team(
    name="Investment Committee",
    members=[bull, bear],
    instructions="Debate the position.",
)

writer = Agent(
    model="openai:gpt-5.4",
    instructions="Write a 200-word investment brief.",
)

workflow = Workflow(
    name="Stock Research",
    steps=[
        Step(name="Research", agent=researcher),
        Step(name="Debate", team=committee),
        Step(name="Report", agent=writer),
    ],
)

workflow.print_response("Analyze NVIDIA for investment.")

Three primitives

Primitive	Use it to
Agent	Build autonomous programs with a model, tools, and instructions.
Team	Coordinate multiple agents to achieve complex tasks.
Workflow	Achieve deterministic output using step-based execution.

Why the Agno SDK

• Just Python. No DSL, no YAML, no custom language to learn.
• Fast. Agents instantiate in microseconds with a minimal memory footprint. You’ll spin up one per request for isolation and permission scoping. Extreme performance is mandatory.
• Agnostic. 30+ model providers, 100+ integrations, 20+ vector dbs, 10+ databases. Use any model provider, any database, any cloud provider.
• Platform-ready on day one. Sessions, memory, knowledge, tracing, evals, and human-in-the-loop come built in. The same agent you prototype in a python script runs on your agent platform behind AgentOS.
• Three primitives. Agents for autonomy, teams for coordination, workflows for deterministic control. Compose them for your use case. Same SDK.

Capabilities

Turn Agents, Teams, and Workflows into production software.

Capability	What it adds
Models	30+ providers behind one API
Tools	100+ integrations and the ability to write your own
Memory	Per-user and per-session memory
Knowledge	RAG over documents, URLs, and databases
Learning	Self-learning agents that improve over time
Reasoning	Built-in chain-of-thought and reasoning agents
Multimodal	Image, audio, and video input and output
Input & Output	Pydantic-typed structured input and output
Context	Add live context to every run
State	Per-agent, per-team, per-workflow state
Dependencies	Inject runtime values into prompts
Skills	Reusable agent capabilities
Hooks	Pre-run and post-run callbacks
Guardrails	Validate input and output
Human-in-the-Loop	Pause runs for approval, input, or external execution
Database	Persist sessions, memory, knowledge, and traces
Sessions	Session lifecycle and metrics
Evals	Test agent quality at scale
Tracing	OpenTelemetry tracing for every run

Other Frameworks

Agno can also serve agents built with the Claude Agent SDK, LangGraph and DSPy through AgentOS. Wrap the external agent in ClaudeAgent, LangGraphAgent or DSPyAgent and register it like any native agent. See Multi-Framework Support.

Where to start

Agents

Single autonomous programs with a model, tools, and instructions.

Teams

Coordinate multiple agents to achieve complex tasks.

Workflows

Achieve deterministic output using step-based execution.