Agno

You will like this if you have an agent that works in a notebook and dread turning it into a service. The daily trait is the runtime: the README's recommended start is a prompt you hand to your coding agent, which clones a starter template and brings up a REST API, a Postgres database and a control plane in Docker. Agents, teams and workflows are the primitives, and Ollama is on the list.

Where it hurts is that you inherit a whole platform when you wanted a library. Pick it for shipping agents as a product. Pick LangGraph if you want the graph and would rather own the server.

The risk is churn. The project was phidata, then Agno, then Agno plus an AgentOS runtime with fifty-plus endpoints, SSE and websockets to keep stable. Every repositioning moves the API under the people who adopted the previous one, and a runtime that promises durable execution and distributed state has far more surface to break than an SDK. The changelog is the only record of what moved.

The consequence: pin versions and read that changelog before every bump. What it does right: human approval is a first-class primitive, so a run pauses for confirmation and tools that need an admin can be blocked rather than trusted.

The architecture is legible. 1. Context: providers fetch live data from Slack, Drive, wikis, MCP and custom sources at run time rather than at index time. 2. State: sessions, memory, knowledge and traces are written to a database the user owns. 3. Observation: OpenTelemetry tracing with run history. 4. Verification: evaluations and simulations are offered as a learning loop, which is asserted rather than demonstrated.

No benchmark is published, and the mechanism behind the durable-execution claim is not described in the README. The observation: a framework that pulls context live at each run has chosen freshness over reproducibility, and does not say so.

Give away the runtime, charge for the window into it. There are no per-event or egress fees; the meter is live connections at $95 a month each, so revenue scales with deployments, not usage. The homepage claims a Fortune 5 customer, SOC 2, UnitedHealth Group and IBM logos, and two million of something a month.

Moat: the runtime's install base, which is only a moat if teams also adopt the hosted control plane. Likely acquirer: a data platform that wants agent runtimes next to its warehouse. Position: long the framework, watch whether the control plane sells.

The demo is an agent answering support tickets in a browser. Pro is $150 a month for three seats and $30 for each seat after that, so sixty engineers is $1,860 a month before the $300 SAML SSO add-on; audit logs and end-user roles are in Enterprise at custom pricing. Telemetry sends an event per agent run by default and is turned off with AGNO_TELEMETRY=false, the first line of any rollout doc.

CI fit is good: it is Python in a container. Onboarding is a week with unusually good documentation. Approved with conditions: Free tier for development, Pro only once SSO is in the budget.

Apache-2.0 and pip install agno, with an agentos-docker template that stands the whole thing up on my own box. Ollama is a supported provider, the storage is a Postgres I run, and AgentOS exposes my agents as an MCP server, so Claude or any other client talks to them without my writing a bridge. The docs themselves ship as an MCP server and as llms-full.txt, which is the right way to feed them to an agent.

The catch: the management UI is a hosted web app that reads from my runtime, and running it myself is an Enterprise add-on. Runtime open, dashboard closed. Fork the runtime; write your own dashboard.