Agno

Agno is an open-source Python framework for building AI agents, teams of agents, and multi-agent workflows that are meant to run in production rather than just a notebook demo. It gives agents memory, knowledge retrieval, tool use, reasoning, and multimodal input (text, images, audio, video) through a single model-agnostic interface that works with more than 30 model providers, so switching between OpenAI, Anthropic, or a local model doesn't mean rewriting the agent. The framework was formerly called Phidata before rebranding to Agno.

The core selling point is speed and low overhead: agents instantiate in microseconds and use a fraction of the memory of heavier orchestration frameworks, which matters when you're spinning up thousands of concurrent sessions rather than one chatbot. Alongside the SDK, Agno ships AgentOS, a FastAPI-based runtime for deploying, monitoring, and managing agent systems in production, with built-in support for state persistence, guardrails, human-in-the-loop review, context compression, and protocols like MCP and A2A.

It's aimed at engineering teams building real products on top of agents, not just prototyping, things like customer support systems, internal automation, or multi-agent research pipelines where latency and resource cost actually matter at scale. Its open-source repo has grown to around 40,000 GitHub stars, making it one of the more visible frameworks in this space alongside LangChain and CrewAI.