Privacy, explained

Your shelf.
Your choices.

This policy explains the difference between using a saved creation and asking AI to make or change one.

Tinker by Ideal Intelligence · Updated September 18, 2026

Using saved applets

Your applet files and their separately saved notes, scores, routines and other state are stored on your device. Opening a saved applet does not automatically send its saved state to an AI model. Tinker does not include advertising, Firebase Analytics or Crashlytics.

Creating and editing with AI

Generation needs an internet connection. In cloud workshop mode, Firebase and Google Cloud process your request, anonymous user identifier, job details and generated applet. Anthropic processes the request and generated content to build or repair it. When you ask for an edit, the current applet HTML is also sent. A rendered preview of generated content may be sent to Anthropic for visual quality checking.

With your own API key, the request goes directly from the app to Anthropic. This still uses an online AI service; it is not an on-device model. The provider's charges and policies apply. The key is retained locally until you remove it in Settings and is used to authenticate requests to Anthropic.

Separate applet state is not automatically included in an edit request. However, personal information you type in a prompt or embed in applet code will be included. Avoid including information you do not want these services to process.

Cloud identity and technical data

The cloud workshop uses Firebase anonymous authentication. Tinker creates this anonymous technical identity automatically the first time you open the app — there is no email address, password or sign-in step anywhere in Tinker. All cloud creation and editing, and all purchasing and restoring of purchases, use that same anonymous identity. The bundled examples and saved offline applets need nothing at all. Firebase Authentication processes this persistent anonymous user ID. The ID links your generation jobs, build wallet, subscription status and usage counters. These records are not anonymous in the sense of being impossible to link together, even though no email address or other personal identifier is ever attached to the ID.

Service providers process technical data such as IP addresses, SDK or platform information and function invocation metadata to operate and protect the service. Firebase SDKs may also process installation or messaging identifiers. Operational logs can contain job status, errors, model token usage and execution metrics. Daily request counters enforce service limits; failed attempts can count toward those limits even when job builds are returned. We do not use these records for advertising or cross-app tracking.

Barcode scanner SDK diagnostics

The barcode scanner uses Google ML Kit. In addition to processing barcode images on the device, the SDK collects device and app information, per-installation identifiers, feature usage events, API configuration, error codes and performance metrics such as latency. Google uses this technical data for SDK diagnostics and usage analytics. This is separate from Firebase Analytics, which Tinker does not integrate. See ML Kit disclosure for Apple platforms and ML Kit disclosure for Android.

Purchases and RevenueCat

Apple or Google processes store payment details. Tinker and RevenueCat process purchase history, product and transaction identifiers, purchase and renewal dates, store, subscription status, refunds, and related technical information to validate purchases, provide access, prevent fraud and understand subscription performance. RevenueCat receives your Tinker anonymous user ID, which we link to your device's Firebase identity. Tinker has no email address to set as a RevenueCat customer attribute — none is ever collected. Neither Tinker nor RevenueCat receives your full payment-card details through this flow.

Restoring purchases (Restore Purchases) links Tinker Pro, and any unused builds from your current Pro period, to a new device through your Apple or Google account. Top-up builds you bought without Pro are not linked across devices this way — they stay only with the device and install where you bought them, because they are consumable purchases.

Our server keeps the authoritative wallet and purchase-grant history. The purchase provider may process app version, platform, country or locale, currency and network metadata as part of its service. We do not configure advertising attribution or use purchases for cross-app advertising tracking. See RevenueCat privacy.

Retention

Cloud job records are normally deleted after your device confirms the result was saved. A scheduled cleanup checks every six hours for jobs older than 24 hours, in batches. An undelivered job will normally remain for about 24–30 hours; outages or a cleanup backlog can extend this. Your anonymous identity, build wallet, subscription state, purchase records and usage counters have separate retention and are not removed merely because a job or local applet is deleted.

Provider logs, backups and Anthropic processing follow their applicable service terms and retention settings. Tinker's job cleanup is not a promise that every provider copy disappears at the same moment.

Phone permissions and external services

Some applets request camera or photo access, location, a picked contact or file, microphone input, speech recognition, step sensors, NFC, or local-network remote control. Sensitive applet capabilities require your consent and applicable system permissions. You can review applet permissions from its menu and revoke OS permissions in your device settings.

Camera images, route points and selected data stay within local applet state unless you choose a transfer. Microphone loudness readings are local; voice-to-text uses your phone's speech recognizer, which may use an online service. Remote-control commands communicate with devices on your local network. Sharing, opening a link, or handing content to another app transfers the content to the service you select. These system handoffs are separate from the WebView's blocked direct network access.

Exports

Share a copy exports the applet HTML without its separate state file. Anything embedded directly in the HTML remains in that copy. An applet's own share or export function can include the content you select. Review the file before sharing it.

Content reports

Settings → Privacy & support contains an in-app report form. It sends the app name or request, your description, your workshop user ID and a timestamp to Ideal Intelligence for content review. It does not automatically attach applet files or saved state. Reports and their rate-limit records are retained for 30 days, then removed by scheduled cleanup. A cloud-data deletion request removes your reports sooner. We limit reports to 10 per cloud identity per UTC day to prevent abuse.

Your controls and deletion requests

Delete an applet from the shelf to remove its local files and saved state. Remove an API key in Settings. Use Settings → Privacy & support to withdraw AI-processing permission or request cloud deletion — no account or sign-in is needed to do this. Cloud deletion removes Tinker workshop jobs, reports, production and test wallets, subscription allowance and usage records, while leaving local applets intact. It can remove your access to unused purchased builds. Restore Purchases works from your store account; contact support if a purchase is associated with a device or install you no longer have.

Deleting the app or its data does not cancel an Apple or Google subscription. Cancel through your store's subscription settings to stop future renewals. Tinker shows a subscription management route; it does not cancel on your behalf.

Cloud deletion also queues removal of the RevenueCat subscriber profile; failed requests are retried by the scheduled six-hour cleanup. Your device's anonymous Firebase identity is deleted directly — there is no password or reauthentication step, because there is no account to reauthenticate.

A minimal operational deletion marker with the user ID, status and timestamps normally remains for 30 days to prevent stale requests from recreating data. Pending deletions are retried. Separate minimal purchase records remain after deletion: hashed account and transaction identifiers, product and store, production/test environment, purchase or expiry dates, and grant/refund status. Hashed event and deletion records prevent duplicate grants and fraudulent replay. These records are pseudonymous, not necessarily anonymous, and currently have no automatic expiry. They are retained for purchase integrity, fraud prevention and accounting. They do not retain your prompts or applet content, and there is no email address to retain in the first place.

For support-message deletion or help with purchases, contact us. We may need a purchase reference that safely identifies the relevant record. Never send an API key, password, full payment-card number or private applet files. Provider logs, backups and store transaction records have their own retention and deletion processes.

Support and children

If you email support, we receive your email address, message and attachments you choose to send. We use them to handle your request. Tinker is intended for adults and is not directed to children. Do not submit children's personal information to AI generation.

Service policies and contact

Firebase privacy · Anthropic privacy

Questions or requests: nalhamzy@gmail.com. We will update this page when Tinker's data practices materially change.