Agno

Agno is an open-source Python framework for building AI agents, paired with AgentOS, a runtime for deploying and managing multi-agent systems in production. Rather than relying on graphs or chains, it takes a plain-Python approach: developers assemble agents from modules for memory, knowledge retrieval, and tool use, then swap in different language models, databases, or vector stores without rewriting application code. The framework is aimed at engineering teams that want to own their agent infrastructure instead of depending on a third-party hosted service. AgentOS adds over 50 API endpoints covering runs, sessions, memory, knowledge, and traces, along with a web console for monitoring agents in production. It supports more than 30 model providers and works alongside other agent frameworks such as LangGraph and the Claude SDK rather than requiring a full rewrite. Production features include stateless execution, per-session isolation, JWT-based role-based access control, approval flows, audit logs, and guardrails for governance. Agno can be deployed to any cloud environment, including airgapped setups, and connects to messaging channels like Slack, Telegram, and WhatsApp, as well as MCP servers. It offers both a free open-source framework and a paid AgentOS runtime. One limitation is that much of the governance and monitoring tooling sits behind the commercial AgentOS product rather than the open-source core.