Keep your tokens safe.
An HTTP MCP token is a credential that grants access to your memory store. Don't expose it in chats, code or public repositories; if you suspect a leak, revoke it immediately and issue a new one.
Terms of service
What to keep in mind
when you use ASR.
These terms describe what the ASR memory store provides and the responsibilities its users take on.
About this translation
This English version is provided for convenience. If it differs from the Korean version, the Korean version governs.
Article 1 · The service
What does ASR provide?
A store that lets you bring back the memories you left, even when your tools change.
ASR is an MCP service that stores, word for word, the text requested by a user or by an AI client acting on the user's behalf, and lets it be searched, recalled, corrected and exported afterwards.
The service does not vouch for an AI client's conversations or answers. What ASR provides is the stored record and the tools to work with it.
Article 2 · Accounts and tokens
You manage access.
Use only your own space.
Tokens and user identifiers are what separate storage spaces. You must not use another person's token or bypass permissions.
Tell your AI the scope, too.
You decide which memories a connected AI may save and search. Stored records are processed according to your requests and settings.
Article 3 · User content
We preserve the original, but don't guarantee it is accurate.
A memory is reference material.
Rights and responsibility for the text you save rest, in principle, with you. When you store other people's personal data, works or confidential information, you must confirm the necessary rights and consent yourself.
Automatic classification, summaries and search can be wrong. Correct a wrong tag with memory_retag, and confirm important decisions against the original and independent sources.
Article 4 · Data processing
What is sent outside, and what is encrypted.
By default, OpenRouter may be used for embeddings and Google Gemini for automatic classification. The original text of regular memories is not end-to-end encrypted; only secret_* values are encrypted separately.
More detail
Where your data goes, and how to turn off or switch outside AI services, is described in Security & data.
Article 5 · Retention and deletion
Deletion happens in two steps.
Delete request
memory_delete removes a memory from search and recall and moves it to the trash. The original remains and can be restored with memory_restore.
Permanent destruction
memory_purge runs only when you explicitly say “permanently delete”. It cannot be undone with the service's tools.
Limits of backups
Even after deletion from the database, data may not be erased everywhere immediately, depending on backup retention periods and outside AI services' policies.
Article 6 · Restrictions and service limits
The following are not allowed.
Unauthorised access
Using another person's token, or bypassing the isolation between storage spaces.
Infringing rights
Collecting or storing other people's personal data, confidential information or works without consent.
Disrupting the service
Excessive repeated calls, malicious input, or abnormal use that causes outages.
Outside dependencies
Outages or policy changes at outside AI services may lower search and classification quality. The scope of storage and retrieval may also change with operating conditions.
Article 7 · Changes and contact
Changes are announced on this page.
We give advance notice before changing tool names or required input formats. Changes needed for operations, law or security are shown on this page with their content and effective date.
Contact
idoweddings@naver.com · We accept requests about accounts, tokens, deletion and your rights.
About this legal document
This page is a notice describing the current facts of how the service operates and its scope of use. Where an individual contract, mandatory law or legal advice applies, that document and professional review take precedence.