Privacy
Captured text stays on disk unless you opt in to a model.
Snibl stores items in a SQLite file under the Electron user data directory. There is no account, no cloud sync, and no telemetry.
Classification opt-in
Sending captured text to a model is off by default. Settings labels this clearly: “Send captured text to a model for classification and chat.”
Captured text goes to the chosen provider only when that toggle is on. When it is off, or when no provider is configured, items save as Unsorted with the reason classification not configured, and the item chat does not send anything. Snibl does not send captured text anywhere else. Favicons are not fetched from the network; source links show the domain and page title.
Local models
You can point Snibl at an OpenAI-compatible endpoint such as Ollama at http://127.0.0.1:11434/v1. Text then stays on the machine that runs the model.
Any provider
Paste a Mastra provider/model id and the API key for that provider. Captured text is sent to that provider only.
The API key stays on this machine. When Electron safeStorage encryption is available, the key is encrypted at rest with the OS secret store. If the OS secret store is missing, Settings says so and the key is stored in the local SQLite file in plaintext.
Snibl never logs the key. It never enables a remote model for you.
License and plan checks
Activating or validating a Pro license sends only the license key and a device name to Dodo Payments’ public license endpoints. Showing device count, update windows, usage, and AI add-on status may also send the license key and license-key id to Snibl’s docs Worker (/billing/status, /billing/portal, /billing/devices). Those routes use a server-side Dodo API key. Captured text, search, chat, and model keys are not included.
There is no account.
Hosted AI
Hosted AI is off until you turn it on in Settings → Plan, after a plain disclosure. It is a Pro add-on only. When it is on and you have not set a BYOK key, classification and Ask AI send captured text to Snibl’s Worker (/ai/classify, /ai/chat) and then to Sarvam AI. The Worker enforces the add-on subscription and a per-license monthly token cap. Pro without the add-on keeps using BYOK only.
App updates
Packaged builds check only https://snibl.app/downloads (latest-mac.yml and a .zip on macOS, latest.yml and the NSIS .exe on Windows). That request does not include captured text, settings, or API keys. No captured data is sent. Development builds (pnpm dev:desktop) do not check for updates. macOS also skips checks unless the running app is Developer ID signed. Packaged Windows builds check even while unsigned.
Custom OpenAI-compatible endpoints
If the Mastra router does not list a provider, set a base URL. Sarvam AI’s Token Factory Chat Completions API is OpenAI-compatible at https://api.sarvam.ai/v2 (Sarvam docs).