# Agent SDK — built for coding agents

- Canonical: https://agno-com-nine.vercel.app/products/sdk

Agno is a Python SDK for building agent platforms. It gives you three primitives: agents, teams and workflows. You attach capabilities like memory, knowledge and learning as your use case needs them. Agno 3.0 is built for coding agents. The docs ship as an MCP server at https://docs.agno.com/mcp. Give your coding agent the MCP command and it assembles your system from documented parts. An agent is one model with tools and instructions. A team is a group of agents that work together in one of four modes: coordinate, route, broadcast and tasks. A workflow runs agents, teams and functions through six step types: Step, Steps, Parallel, Loop, Condition and Router. Capabilities include learning, context providers, memory, knowledge, state, storage, sessions, guardrails, hooks, dependencies, a filesystem and human in the loop. There are 30+ model providers and 100+ tool integrations behind one API. For operations there is background execution, evals, observability and a scheduler. An agent is created in 3.2 microseconds and holds 5.2 KiB of memory, measured on the public harness you can run yourself. Register the same agents, teams and workflows with AgentOS to run them behind a FastAPI application.