trakd

Privacy for the
Trakd catalog plugin.

Effective November 6, 2026. This notice supplements the Trakd privacy policy for the catalog service operated by Roji Labs LLC.

Requests and catalog information

When ChatGPT calls a Trakd tool, this service receives the search words, public recording link, entity ID or other parameters needed for that request. It uses them to retrieve public catalog information and source links. Search results may be cached in server memory for up to 30 seconds.

This version does not require Trakd sign-in. It does not access your private Trakd library, member activity, audio uploads or full ChatGPT conversation history. Catalog requests do not invoke audio-recognition providers or change catalog records.

Returned public records may include artist names, biographies, press photographs, country and genre information, public artist/store links, recorded set and track credits, cue times, and public venue names and business addresses. We retrieve these catalog fields to answer the requested music question. Private member records are excluded.

Security and operations

Request information is processed by our hosting and database providers, Vercel and Supabase, to operate the service. The service uses a public database key for allowlisted reads and a separate server-only capability for the fixed catalog gateway and bounded operational activity logging. It cannot change catalog records. The gateway maintains shared request and returned-row counters using a token-keyed digest of the network address; raw addresses and prompt text are not stored in these counters. Quota counters older than 48 hours are cleaned on subsequent admission requests. No scheduled cleanup runs when the service is idle. Network quotas may be shared by users whose requests are proxied through the same address. Hosting providers may process network addresses and operational request logs under their policies. The service does not deliberately log tool-request bodies or return these operational details in answers.

Plugin activity

For service monitoring, we store the time, protocol action or tool name, outcome, response duration, returned-record count, plugin version and up to three resolved public catalog references. We do not store raw search text, request parameters, prompts, full conversation history, account identities, network addresses or network hashes in this activity feed. Connection handshakes are anonymous protocol events, not a list of users or chats. These records are restricted to workspace owners and retained for 30 days; older records are deleted when subsequent activity arrives. A monitoring outage does not block catalog answers. We do not use these records to train models.

Questions and corrections

Contact support@trakthat.app for privacy questions or corrections to public catalog information. ChatGPT’s handling of your conversation is governed separately by OpenAI’s policies.