Tetrees AI Privacy Policy
Last updated: 2026-08-24
1. Data we process
We process account/profile data; authentication and security logs; purchases, licenses, Points, seller payouts and disputes; Pack listings, TAIP metadata, images and Agent AVCP evidence; prompts, model choice, usage totals, tool-call metadata and run results; encrypted approved memories and growth history; support communications; first-party analytics and device/network data. Stripe handles full card data; Tetrees does not store card numbers.
2. Agent and model processing
For a hosted or BYOK run, the selected provider (OpenAI, Anthropic or Z.AI) receives the prompt and the minimum Pack context, approved memory, conversation history and tool result needed to answer. Its own terms and retention rules apply. BYOK credentials are used only for the request and are not persisted in Tetrees databases or logs.
If you attach a supported file to a run, Tetrees validates and extracts bounded text in request memory. The raw file and extracted text are discarded when that request finishes and are not written to Pack storage or run history. We retain bounded metadata such as filename, media type, size, digest and truncation state. The selected model provider may receive extracted content when the Pack uses its attachment-reading skill, and the resulting agent answer is stored like any other run result.
The hosted cited-web-search skill sends the user's bounded search query to the configured search/model provider. Search text and URLs can appear in the run result and skill-call evidence. Tools, resources, prompts or APIs configured in a user's own MCP host are processed under that host's privacy terms and permissions, not as Tetrees-hosted extensions.
We store run status, model, token usage, Points settlement, a prompt digest, result and growth proposals to provide history, billing, safety, support and abuse prevention. Approved memories are encrypted at rest, separated by user and Pack, and not included in another user's Pack instance.
3. Why and with whom
We use data to operate accounts, exchange and entitlements; execute agents; settle Points, tax and payouts; perform Agent AVCP; prevent abuse; support users; resolve disputes; meet legal duties; and improve reliability. Processors include AWS/S3, Stripe, Elastic Email and the model provider you select. Sellers receive only the buyer and transaction information reasonably required for support and disputes.
4. Retention, transfer and security
Account and runtime data is retained while needed for the service. Financial, audit and dispute records may be retained longer by law. Revoked tokens cease access immediately. Artifacts use access-controlled storage and short-lived downloads; credentials are hashed or encrypted; approved memory uses authenticated encryption. No security method is perfect.
Tetrees operates from Korea and providers may process data in other countries, including the United States. We use available contractual and technical safeguards.
5. Choices and rights
You can avoid hosted model billing by using BYOK, but the chosen provider still processes the request. Do not put unnecessary personal or confidential data into prompts or Packs. Subject to law, you may access, correct, export or delete account data and object to or restrict processing. Marketing emails include unsubscribe controls; necessary security, transaction and support mail remains. Contact business@insightvessel.io.