Privacy

Inference stays local

Chat and embeddings run through models you host with Ollama (or other local backends the app is configured to use). Prompt text and retrieved document chunks used for generation are not sent to AnyLM cloud inference — there is no hosted model API in the product path.

What lives on your device

Projects, threads, messages, attached references, embeddings, generated artifacts, and settings are stored in the app's local data directory on your computer. Uninstalling the app or clearing that directory removes them.

Accounts and cloud services

Sign-in (email/password, Google, GitHub) uses Firebase Authentication. Organisation membership, policy, and cooperative usage records may be stored in Firebase Firestore so limits and roles can sync across devices for the signed-in account. Firebase processes that account data under Google's terms for the project that hosts AnyLM.

Session tokens on the device are kept in the OS keystore when available (Keychain / DPAPI / libsecret), with a restricted file fallback only where no keyring exists.

Website and downloads

The marketing site at anylm.app serves product pages and resolves download links from public GitHub Releases. Standard web server and CDN logs may include IP address, user agent, and requested URLs. We do not use the site to run third-party advertising pixels.

Updates

When auto-update is enabled, the desktop app checks GitHub Releases for new installers. That check reveals that a machine requested update metadata; it does not upload chat content.

Contact

Privacy or security questions: yasharyan307@outlook.com. Security reports should follow SECURITY.md.