
agno
Python agent framework plus a FastAPI runtime you host yourself; the management UI is Agno's own hosted product.
What is agno?
Agno is three layers you can adopt one at a time: a Python SDK for agents, teams and workflows; AgentOS, a FastAPI service that turns them into a REST API with sessions, JWT-scoped access, scheduling and OpenTelemetry traces written to a database you own; and a web control plane for watching runs. The first two are Apache-2.0 and run wherever containers run; apart from a usage ping you can switch off, nothing has to reach Agno. The control plane is the part that does not follow that rule — free against a runtime on your own machine, and a paid plan to point it at a deployed one. The project also moves quickly, with breaking changes in most minor releases, so pin a version before you build on it.
What can you do with agno?
- An agent is a model, a set of tools and instructions — Twenty lines give you a working agent: choose a model, hand it tools such as a working folder or web search, write the instructions as prose, and run it. Responses stream to the terminal while you iterate on the wording.
- Give it a database and the conversation survives the process — Attaching a database turns on sessions: runs grouped under a session ID and tied to a user ID, so earlier messages can be replayed into the next request. Memory goes further and extracts durable facts about each user that follow them into later conversations.
- A knowledge base the agent decides when to search — Load files, URLs or raw text into a knowledge base backed by one of the supported vector databases. The default is what the docs call Agentic RAG — the agent searches when it judges it needs to, rather than context being injected on every turn — and agents can write new insights back into it.
- Teams delegate, workflows follow a fixed path — A team has a leader that coordinates, routes or broadcasts work to members, each with its own role and tools. A workflow is the opposite: steps in a set order with loops, conditions and parallel branches. Agno's own docs are candid that a team costs extra model calls, latency and tokens.
- AgentOS turns the agents you wrote into a service — Wrap them in AgentOS and you get a FastAPI app: a run endpoint per agent, team and workflow, streaming responses, JWT-checked permissions, a cron scheduler inside the same process, and execution traces written to your own database. Slack, Telegram and an MCP server are switches, not projects.
- Context Providers keep the tool list short — Rather than mounting a dozen Slack tools next to a dozen Drive tools, each external system sits behind one read tool and one optional write tool, with a small sub-agent doing the actual calls. Agno's docs give the reason plainly: past twenty tools, models start picking the wrong one.
Before you choose agno
- The SDK and AgentOS are Apache-2.0, but the control plane is Agno's hosted product: as of August 2026, free against an AgentOS on your machine, $150/month to connect a deployed one, self-hosted only on Enterprise.
- Upgrades cost real work: v1 to v2 needed a database migration script, 8 of the 10 v2 minor releases through v2.9.0 shipped breaking changes, and a v3 alpha was tagged in August 2026.
Star history
17 Aug to 30 Oct · +1.2k
Frequently asked questions
Is agno free for commercial use?
agno is released under the Apache-2.0 licence — OSI-approved open source, which permits commercial use.
How can agno be deployed?
agno is available as Self-hosted / Runs locally.