SWE-agent is an open-source (MIT) agent for autonomous software engineering. A language model, for example GPT-4o or Claude Sonnet, is given tools to resolve GitHub issues on its own, look for vulnerabilities in software or carry out other tasks. You install it yourself, from source or in the browser, and connect a model provider with your own keys. It is an open alternative to commercial autonomous coding agents such as Devin.
According to the project page, SWE-agent is built and maintained by researchers from Princeton University and Stanford University, both in the US. No commercial company stands behind the project, and there is no vendor cloud: it runs on your own infrastructure. The project states that it is in maintenance mode and points to its successor, mini-swe-agent, for new development. What leaves your systems depends on the model provider you configure; a European organisation therefore decides for itself where source code and prompts end up.