Agno Documentation
by Agno
Python multi-agent framework plus a FastAPI runtime you deploy into your own cloud.
Overview
Agno, formerly Phidata, is a Python framework and runtime for multi-agent systems at roughly 42,000 GitHub stars and 5,800 forks. It markets itself on being lightweight: the March 2025 Hacker News post that first drew attention was titled 'Agent framework 10,000x faster than LangChain', a claim about how quickly an agent object instantiates and how little memory it occupies rather than about response latency, which in any real deployment is dominated by the model call. The product is three layers. The SDK covers agents, teams — multiple agents with delegation — and workflows with sequential, parallel and human-in-the-loop steps, plus more than 100 toolkits, MCP support, knowledge and RAG over vector databases, session memory and chat history, structured and multimodal input/output, over 30 model providers and over 20 database backends. AgentOS turns that into a stateless FastAPI service exposing more than 50 endpoints and an MCP server, with durable execution, distributed state, JWT-based role-based access control, and OpenTelemetry tracing. The Control Plane is a web UI for traces, sessions, evals, metrics and audit logs. The positioning is explicitly bring-your-own-cloud: the runtime is designed to deploy into infrastructure you control — Docker, Kubernetes, AWS, GCP, Azure, Railway, Fly, Modal or Render — so agent data and traces never leave it.
At a Glance
- Topic
- Frameworks
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Reference documentation, self-paced — about 3 hours to work through agents, teams, workflows and AgentOS
- Provider
- Agno
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Compose a Python agent from a model, instructions, tools, memory and a knowledge base
- ✓Build teams where a lead agent delegates subtasks to specialised member agents
- ✓Define workflows with sequential, parallel and human-approval steps for deterministic control
- ✓Attach knowledge from vector databases and choose among 20-plus storage backends for sessions
- ✓Connect MCP servers and any of the 100-plus built-in toolkits as agent capabilities
- ✓Deploy agents behind AgentOS as a FastAPI service exposing 50-plus endpoints and an MCP server
- ✓Instrument runs with OpenTelemetry tracing and enforce JWT role-based access control in production
Highlights
- •Ships a production runtime, not just a library — AgentOS is the deployment story most frameworks omit
- •Bring-your-own-cloud by design, so sessions, traces and agent data stay in your infrastructure
- •Over 30 model providers and 20 database backends, so neither choice is locked in at the start
- •Built-in OpenTelemetry tracing and JWT role-based access control rather than bolted-on afterwards
- •Roughly 42,000 GitHub stars with repeated well-received Show HN launches since March 2025
Who It’s For
Best For
- ✓Python teams building multi-agent systems that need to run in their own cloud
- ✓Engineers who need an agent runtime with auth, tracing and durable execution out of the box
- ✓Developers comparing agent frameworks on operational maturity rather than API ergonomics alone
Prerequisites
- •Working Python and familiarity with FastAPI or a comparable async web framework
- •Understanding of LLM tool calling, prompts and structured output
- •An API key for at least one supported model provider, plus a database for session persistence
FAQ
What is Agno Documentation?
The official documentation for Agno, a Python framework and runtime for multi-agent systems. It covers agents, teams with delegation, and workflows with human-in-the-loop steps, alongside 100-plus toolkits, MCP support, knowledge and memory backends, and AgentOS — a FastAPI runtime that turns your agents into a secure API and MCP server deployed inside your own infrastructure.
Is Agno Documentation free?
Agno Documentation is free to access.
What level is Agno Documentation for?
Agno Documentation is aimed at a intermediate audience. Recommended background: Working Python and familiarity with FastAPI or a comparable async web framework, Understanding of LLM tool calling, prompts and structured output, An API key for at least one supported model provider, plus a database for session persistence.
How long does Agno Documentation take?
Expect roughly Reference documentation, self-paced — about 3 hours to work through agents, teams, workflows and AgentOS. Most learners work through it at their own pace.
What will I learn from Agno Documentation?
You'll learn: Compose a Python agent from a model, instructions, tools, memory and a knowledge base; Build teams where a lead agent delegates subtasks to specialised member agents; Define workflows with sequential, parallel and human-approval steps for deterministic control; Attach knowledge from vector databases and choose among 20-plus storage backends for sessions; Connect MCP servers and any of the 100-plus built-in toolkits as agent capabilities; Deploy agents behind AgentOS as a FastAPI service exposing 50-plus endpoints and an MCP server; Instrument runs with OpenTelemetry tracing and enforce JWT role-based access control in production.
Topics
Sources
This page was written from 3 sources, 2 on domains other than docs.agno.com.