Privacy notice

Effective August 14, 2026. This notice explains what OnMe collects, why it is used, where it may be processed, and the choices available to you.

Information collected

OnMe may collect account and contact information; Stripe customer, subscription, invoice, and entitlement state; confirmed projects and jar permissions; task definitions, schedules, destinations, timezones, and success conditions; captures, screenshots, article text, submitted artifacts, evidence references and hashes; AI-generated classifications, verdicts, reasons, confidence and credibility estimates; acceptance, reversal, outcome, and sweep history; API-key identifiers and scopes; and device, request, delivery, error, security, and usage logs.

Sources

Information comes from you, your browser or device, the execution host and tools you choose to connect, service providers acting for OnMe, and payment events from Stripe. A connection configured inside ChatGPT, Claude, Cursor, Telegram, or another host remains governed by that provider; OnMe receives only what the host or you submits to OnMe.

How information is used

OnMe uses information to authenticate customers, provide and secure the service, enforce jar and confirmed-project permissions, evaluate captures and scheduled evidence, decide whether to interrupt, deliver requested results, meter entitlements and model costs, process billing, respond to support requests, prevent abuse, debug failures, and improve your decision history from feedback and outcomes.

AI processing

On the Sieve plan, OnMe makes zero calls using OnMe provider credentials; content sent to a customer-controlled execution host is processed under that customer’s separate provider account and terms. Where supported and separately certified, an iPhone may evaluate a bounded capture on device. Compact same-account, same-Track feedback preferences and content-free numeric training examples remain in that iPhone’s Keychain. Feedback waiting for server synchronization is held separately in a bounded Keychain-protected queue for up to 30 days; a reversal reason may be present there so the requested decision trace can be completed, but reasons never enter the private learning profile. After an untouched chronological holdout proves a personal ranker improves on the deterministic baseline, the app may train and use that bounded ranker locally; its examples and parameters are not uploaded. Only a decision-specific request to suppress an otherwise-eligible threshold interruption is submitted, and server policy decides whether it is permitted. On the Hosted plan, submitted content may be sent to the model and search providers listed in the AI/provider disclosure to classify, summarize, compare, or generate a verdict. OnMe routes deterministic checks first and uses model processing only where the product flow calls for it. Do not submit secrets or sensitive personal information unless it is necessary and you are authorized to do so. Protect reminders should use minimum timing details; personal details are not permitted to grant project-reasoning access.

Service providers and disclosures

Depending on the feature used, OnMe relies on providers such as Supabase for authentication and data storage, Railway for hosting, Upstash for delivery jobs, Anthropic and OpenAI for model or embedding processing, Stripe for payments, Telegram and Expo for delivery, and Resend for email. These providers receive information only for the relevant service function. OnMe may also disclose information when you direct it, to protect rights and security, during a business transaction subject to appropriate safeguards, or when legally required.

Payments

Stripe collects and processes payment-method details on its hosted Checkout surface. OnMe receives customer, subscription, payment-status, and limited billing metadata needed to provide access; OnMe does not receive or store your complete card number or security code.

Browser storage and logs

The private Tasks screen may keep the scoped OnMe API key in session storage so the screen can reconnect during the current browser-tab session. Close the tab and revoke the key after using a shared device. The Review screen submits keys in a form body rather than placing them in new URLs; legacy key-in-URL links remain temporarily compatible and should not be shared. OnMe keeps limited redacted request and operational logs to secure and diagnose the service. The public site does not currently use advertising cookies or cross-site behavioral advertising.

Sharing and sale

OnMe does not sell personal information and does not share it for cross-context behavioral advertising. OnMe does not use your content to train a general-purpose model. Information is disclosed only as described above, when you request a destination or integration, or when law requires it.

Retention and sweeping

Resolved items leave active context immediately. Raw content marked as noise is scheduled for compaction after 14 days. Raw content for explored or acted items is scheduled for compaction after 90 days while a minimal decision trace may remain for your personal learning history. Pending encrypted iPhone feedback expires locally after 30 days and can be retried or discarded from the private-learning status screen. Reset private learning removes the device profiles and retains only a content-free reset timestamp in Keychain, preventing earlier server labels from entering that iPhone’s new learning epoch after sign-out. Operational serve and notification logs are scheduled for removal after 90 days; stock snapshots after 180 days; and completed personal calendar details after 30 days. Billing, fraud-prevention, legal, and backup records may remain longer where reasonably necessary. Deletion from backups may take additional time.

Your choices and privacy requests

You can correct task and project permissions, give decision feedback, reset private learning on an iPhone, hard-delete individual items, revoke API keys, pause or delete tasks, and cancel the subscription. A device reset excludes earlier feedback from future local learning but does not delete the separate server decision history. You may request access, correction, export, or deletion of account information by emailing [email protected] from the subscription email. OnMe may verify your identity before completing a request and may retain information where required for security, billing, legal obligations, or dispute resolution. If applicable law gives you additional rights—such as rights to know, delete, correct, limit, or receive equal service for exercising privacy rights—OnMe will honor them.

Security

OnMe uses measures including authenticated access, row-level database policies, hashed API keys, scoped credentials, rate limiting, signature-verified webhooks, idempotency controls, and restricted server credentials. Security depends in part on the execution host and destination you choose. No internet service can guarantee perfect security.

Children

OnMe is built for adult solo builders and is not directed to children under 13. The paid service requires users to be at least 18. Do not submit information about a child unless you have authority and the information is necessary for a permitted reminder.

International processing and changes

OnMe and its providers may process information in the United States and other locations where they operate. Privacy protections may differ by location. Material changes to this notice will be posted here with a new effective date and, when practical, communicated to active subscribers.

Contact

Privacy questions and requests: [email protected]. The responsible operator is OnMe; a legal entity name and mailing address will be added when the operating entity is finalized.