Agno iconAgno

Open-source Python framework for agents, teams and workflows, plus AgentOS: a runtime and control plane you run in your own cloud

43.1k+ GitHub Stars
v3.1.12 Latest (Oct 29, 2026)
$0 Free Plan

Overview

Agno is an open-source (Apache-2.0) Python framework for building agents, multi-agent teams and workflows. The project started as Phidata and rebranded to Agno in January 2025. In September 2025 it added AgentOS, a production runtime that serves your agents, teams and workflows over a REST API and runs in your own cloud, so agent data stays in your infrastructure. AgentOS can also expose agents inside Claude and ChatGPT through MCP, and in Slack and Discord. On top sits a control plane with sessions, traces, memory, knowledge, evaluations, human approvals and a scheduler, plus role-based access control on paid plans. Agno's pitch is speed and a light footprint: it is pure Python, quick to start, and the company publishes benchmarks claiming much faster agent startup and lower memory than LangGraph; those are vendor numbers we could not verify. The company is based in New York and raised a $5.4M seed in August 2024. Main alternatives are LangGraph, CrewAI and Google ADK.

The Verdict

Who Should Use Agno?

Best For

  • Python teams that want to ship agents fast with little boilerplate
  • Companies that must keep agent data inside their own cloud
  • Teams that want a ready control plane for traces, evals and approvals

Not Ideal For

  • TypeScript, Go or Java shops (Python only)
  • Teams that need fine-grained, explicit control of every agent step

What's Great

  • Lightweight, pure Python and quick to get started
  • Runs in your own cloud with no serving fees on any plan
  • Large, active open-source project with frequent releases

Watch Out For

  • High-level abstractions can make it harder to trace why an agent did something
  • Speed and memory claims are vendor benchmarks
  • RBAC and multiple connections need the paid Pro plan

Pricing

View all features & details

Key Features

  • Agents, multi-agent teams and workflows
  • AgentOS runtime served over REST
  • Serve agents in Claude and ChatGPT via MCP, plus Slack and Discord
  • Control plane: sessions, traces, memory, knowledge, evals
  • Human approvals and a scheduler
  • RBAC on paid plans

Company

  • Formerly Phidata; renamed Agno in January 2025
  • AgentOS launched September 2025
  • $5.4M seed (Aug 2024); based in New York

How It Compares

Feature Agno LangGraph Google ADK
Languages Python Python, TypeScript Python, TS, Go, Java, Kotlin
Style High-level agents and teams Low-level graph state machines Agents plus graph workflows
Runtime AgentOS in your cloud LangChain hosted or self-hosted Agent Runtime, Cloud Run, GKE
License Apache-2.0 MIT Apache-2.0

Sources: Agno · Pricing (fetched Nov 2026) · GitHub · AgentOS launch (Sep 2025) · ZenML comparison

Is this your tool? Add a badge