Privacy Policy

Privacy Policy

1. Information the application stores

We store workspace identifiers, usernames, salted password hashes, hashes of session tokens and recovery and API keys, key limits, model selections, request identifiers, timestamps, token counts, status, charges and financial ledger entries. Payment records include invoice references, public blockchain addresses, amounts and transaction details used to verify credit and refunds. Requests submitted through the workspace store their subject, messages, reference, status and operator replies.

From 19 October 2026, accepted API and website chat requests are logged before forwarding to the model provider. This includes the full message history supplied in the request (user, system, assistant and tool messages), tool definitions and structured output instructions, together with the account, API-key identifier, model, request identifier and timestamp. Prompt content is encrypted in the application database and can be decrypted by authenticated administrators. Each administrator content view is audited. New model responses are not separately collected by this logger, but replies included as conversation history in a later chat request are logged as input. The website chat keeps the current conversation in page memory; reloading clears the displayed conversation. HTTP authorization headers, cookies and API credentials are excluded, but any secrets or personal information typed inside messages are still part of their content. Do not submit sensitive data. Content you deliberately include in a private request is stored as part of that request. Administrative access uses identity information supplied by the hosting authentication service.

2. How information is used

Prompt logs are used for service diagnostics, abuse investigation and operational review. Information is used to authenticate access, provide inference, apply limits, calculate charges, verify payments, reconcile failures, handle requests, investigate abuse and maintain security and operational records. Do not put secrets, credentials or unnecessary sensitive information in prompts or private requests.

3. Model providers and infrastructure

Generation content is transmitted to the selected model provider to produce a response. The currently connected route uses NanoGPT and its configured upstream provider. Their processing and retention practices are separate from our application database and are not overridden by this policy. Review the NanoGPT privacy policy before submitting sensitive information.

Hosting and network services, including OpenAI Sites, Cloudflare, the server operator and blockchain RPC providers, may process connection information and operational logs under their own practices. Their logging and retention practices are separate from the application prompt-log retention described here. Cryptocurrency transactions are public; a private workspace does not make blockchain activity anonymous.

4. Browser storage and security controls

Username accounts use a first-party, HttpOnly session cookie that expires after up to 14 days. Session tokens are hashed in the database. Signing out revokes the current session; changing a password invalidates previous sessions. Passwords are stored as salted scrypt hashes, not readable passwords. Key-only access uses sessionStorage for the recovery key in the current browser tab. It sends that key to the service for authenticated workspace operations. Closing the tab or clearing storage can remove this local copy, so save a recovery copy before leaving. Essential hosting authentication or security storage may apply separately. The application does not embed advertising pixels or third-party analytics trackers.

Workspace creation and authentication use short-lived salted, one-way identifiers derived from the connecting IP address or username for rate limiting. Raw IP addresses are not stored in the application database. Hosting and security infrastructure can still process network identifiers. Credentials are hashed or encrypted according to their purpose, access is checked server-side and encrypted payment-system backups are maintained. No system can guarantee absolute security.

5. Retention and deletion

Prompt logs expire 30 days after collection and are no longer available through the administrator interface. Expired rows are deleted when new accepted requests or administrator log reads run cleanup; if the service is idle, physical deletion can occur later. Administrators can delete an individual prompt log earlier, and you may request deletion through your account. Usage and billing metadata are retained separately. Existing encrypted backups may retain copies until their normal expiry; deleting a live log does not delete every backup immediately.

Workspace, usage, billing and private-request records are retained while needed to operate the account, reconcile transactions, investigate disputes and meet applicable obligations. We do not promise a single fixed deletion period for every record. A deletion request may require preserving a limited billing, security or dispute record. Encrypted backups are separate copies; removal from the active application does not imply immediate removal from every backup or public blockchain.

6. Your choices and requests

You can revoke API keys and export request history from your workspace. Use a Privacy & data request to ask for access, correction, account closure or deletion. We may verify ownership before acting. Responses and any explanation of retained records are provided in the same private request. Rights and permitted retention depend on applicable law; this policy does not restrict rights that cannot be waived.