goose

GDPR overview

The site makes no GDPR claim for goose. Across the pages collected, the single occurrence of the word GDPR describes Scaleway, one of the third-party LLM providers listed as a supported endpoint: it commits Scaleway, not goose. No privacy policy, no terms of service and no legal notice exist on the domain, and a sitemap of 317 URLs confirms these pages are absent rather than merely hard to find. There is no Article 27 representative, no data protection officer, no data processing agreement, no subprocessor list and no contact email address of any kind. What exists is architectural rather than contractual: goose runs locally, conversations stay on the machine, and the only publisher-side collection is anonymous telemetry you must explicitly opt into at first launch.

Who owns the data?

goose runs on your own machine, and the documentation states that all conversations and interactions, in both the CLI and the desktop app, are stored locally. In practice you hold everything: session history, generated files and the configuration in ~/.config/goose/. The publisher applies no ownership clause to any of it, for a simple reason - no terms of service and no privacy policy are published on the domain, so no contractual text defines ownership, licence or reuse. The only publisher-side collection is an optional anonymous telemetry stream. Anything you send to the LLM provider you connect falls under that provider's own policy, not goose's.

Reuse rights

No terms of service exist on the site, so nothing restricts what you do with your sessions, your outputs or the files goose produces: they sit on your machine and are yours to reuse without asking anyone. The Apache 2.0 licence covers the software itself, not the content you generate with it, and it also allows you to fork and redistribute goose. On the publisher's side, the only documented collection is anonymous telemetry, opt-in at first launch, whose stated purpose is product improvement. Six categories are listed: operating system, version and architecture; goose version and installation method; provider and model in use; extension names and tool usage counters; session metrics such as duration, number of interactions and token counts; and error types such as rate_limit or auth, without detail. It excludes conversations, code, tool arguments, error messages and personal data, and can be switched off at any time. The documentation warns separately that, depending on the LLM you use, your prompts and the information goose accesses may be sent to that provider and fall under its own retention and privacy policies.

Data retention & training

Hosting summary

There is no publisher-side hosting to describe: goose runs on the user's own machine. The documentation states that all conversations and interactions, in both the CLI and the desktop app, are stored locally, and configuration lives in ~/.config/goose/ through config.yaml and profiles.yaml. No hosting country, region or jurisdiction appears anywhere on the site, which follows from a locally executed tool rather than from an omission. The documentation site itself is served by GitHub Pages behind Fastly, on an anycast address routed from the United States, but that concerns the docs, not user data. Two paths do send data off the machine. First, every request to the LLM provider you configure, which falls under that provider's own jurisdiction and policies. Second, two optional modes: a remote instance started with goose serve, hosted on infrastructure you choose and protected by TLS and a shared secret, and mobile access, which the publisher describes as passing through a Cloudflare tunnel when you enable it.