AutoGen tops three agent frameworks on GitHub stars and ranks last on monthly PyPI installs

Stars only accumulate. The post says a star count reflects bookmarks made when the hype peaked, for AutoGen during its v0.2 launch period in 2024, and that they are never removed when a project goes quiet ^[10]. Installs run the other way, because each CI run or image rebuild that pulls the package is a recurring production cost ^[11].

CrewAI sits between the two on installs, at about 20.7 million a month ^[6]. Divide the post's install figures by its star counts and the gap widens. LangGraph comes to about 1,780 monthly installs per star ^[16]. AutoGen comes to about 18 ^[17]. That is a difference of roughly 100 times ^[18]. The post turns this into a heuristic: a high star-to-install ratio is a freshness signal in reverse, a mark of a project that was once exciting and is no longer relied on in production ^[23].

The ratio is a proxy, and the post says what the install counts contain. PyPI counts include CI pipelines and container rebuilds, so they are not user counts ^[12]. LangGraph is also pulled in transitively by the wider LangChain ecosystem ^[13]. Some of its volume therefore comes from projects that depend on it without having chosen it. For the 65x gap to transfer to your own shortlist ^[5], most of that volume would have to be deliberate, direct installs. The post asks readers to treat the figure as a relative signal, not an absolute ^[12].

The README is the better primary source. Its banner is addressed to new users and directs them elsewhere ^[4]. For code already written on AutoGen, the post names AG2, a fork and not a rebrand ^[8].

The post sorts the three by how they fail at scale: LangGraph on checkpoint growth, CrewAI on prompt-mediated delegation, AutoGen on maintenance ^[20]. A team can size the first two before it commits. The cheapest option is also on the list. If the workflow is three LLM calls in a loop, the post says to use no framework at all ^[22].

Everything here is one author's snapshot. The post says it took its facts from GitHub's REST API, PyPI's release history, the official READMEs, migration guides from the vendors and public issue trackers, and checked them all on 2026-08-16 against Microsoft Agent Framework 1.0, AutoGen 0.7.5, CrewAI 1.15.16 and LangGraph 1.2.11 ^[14]. It warns that the category moves within weeks and says to check those versions on PyPI before acting ^[15]. The author wrote: "Don't pick on star counts. Pick on what shipped, what broke, and what's still being maintained." ^[24]