MultiVAC Privacy Policy
Draft prepared: October 11, 2026
Effective date: October 11, 2026
Operator: Intracoast Network Services
Contact person: [redacted]
Contact: [email protected]
Intracoast Network Services (“we,” “us,” or “our”) operates the MultiVAC Discord bot. This Policy describes information processed by MultiVAC, including information received through Discord and information sent to external providers. It applies to the bot and its associated features. Discord and external services also process information under their own policies.
1. Information we process
Depending on the enabled features and your interactions, MultiVAC processes:
- Discord identifiers and context: user IDs, display names, server and channel IDs, message IDs, timestamps, reply relationships, and permissions needed to handle or authorize requests.
- Requests and conversation content: messages addressed to the bot, its responses, replied-to messages, and a bounded window of nearby channel messages. Nearby messages can include people who did not directly address the bot.
- Attachments and tool inputs: images, videos, supported files, attachment URLs, links you ask us to open, search queries, and locations you provide for weather or other location-based requests. A saved location may persist for later requests; it is not continuous device location tracking.
- Personalization and memory: remembered facts, saved instructions, preferences, conversation persona settings, and automatically distilled summaries of a user's indexed interactions where that feature is available.
- Operational records: provider/model names, request status, tool activity, timing, errors, public evidence links, token counts, estimated costs, attribution to a user, and feature approval or change-history records.
- Support information: information you provide when asking for help, reporting abuse, or making a privacy request.
Do not send passwords, access tokens, payment card details, private medical records, or other sensitive personal information. MultiVAC does not need your Discord password to operate.
2. Why we process information
We use information to answer requests, analyze or generate media, retrieve information, maintain conversation context, apply saved preferences, provide requested memory features, manage usage and costs, enforce permissions and safety controls, diagnose failures, respond to support and privacy requests, and meet applicable legal obligations.
Where privacy law requires a legal basis, we rely on the applicable basis for each purpose, such as providing a service you request, legitimate interests in secure operation, consent where required, or compliance with law. A server administrator's installation of the bot does not replace any individual consent required by applicable law.
3. Conversation context, memory, and reflection
When invoked by a mention or reply, MultiVAC can read up to 24 preceding messages in the same channel, subject to a 12,000-character limit. Display names and user IDs help distinguish speakers. This live context window is processed temporarily; it does not itself create another transcript archive.
Separately, when the Elasticsearch/OpenSearch memory backend is configured, messages addressed to the bot and bot replies can be indexed persistently with their content and Discord identifiers. Referenced messages may also be indexed. Members with the required moderation permission can explicitly index recent channel history, including messages from people who did not invoke the bot. Personalization can retain user facts, instructions, and a distilled profile. Some personalization is keyed by user ID and may apply in later conversations; conversation context and persona settings use their own channel or conversation scope.
If reflection is enabled by the operator, invoking the bot opens a bounded interaction-analysis session. Defaults include a ten-minute lookback, continued processing of messages in that channel while the session is active, and closure after five minutes without activity. Other participants' messages may supply context. Temporary text can be sent to the configured AI provider. The reflection store retains derived observations, hashed actor references, evidence/message identifiers, improvement ideas, and operational or budget records, rather than a second raw conversation archive. Hashed references are pseudonymous and are not necessarily anonymous.
4. Who receives information
- Discord and channel participants: responses are sent through Discord. People who can access the destination channel can normally see them. Direct messages still involve Discord and relevant processing providers.
- AI and media providers: configured features may send prompts, relevant context, saved personalization, attachments, and requested generation or analysis inputs to OpenAI, Google/Gemini, Anthropic, or other configured media providers. Context may include Discord display names or identifiers.
- Search and information providers: relevant queries, image data or public image URLs, locations, and other tool inputs may be sent to services such as Google Search, Google Cloud Vision, SerpApi, OpenWeather, or Finnhub when their feature is used. Opening a supplied link involves the destination website.
- Hosting and storage providers and authorized operators: these parties may process stored information or access it when necessary to maintain the Service, investigate incidents, or fulfill requests.
- Legal or safety recipients: we may disclose information where required by law or reasonably necessary to address unlawful activity or protect rights and safety, subject to applicable law.
Server members with the Discord Manage Messages permission can view per-user
usage rankings through the usage command, including user IDs, call counts,
and estimated costs. This is distinct from public status statistics,
which contain aggregate call and token counts without user IDs or prompts.
We do not sell Discord API data or share it with advertising networks or data brokers. MultiVAC does not operate a model-training pipeline. Sending content to a provider for a response is still disclosure to that provider; its retention and permitted uses depend on its applicable service terms and account settings. This Policy does not promise zero provider retention or that every provider processes data identically. Relevant provider notices include Discord, OpenAI, Google, and Anthropic.
5. Retention and deletion
Temporary conversation windows and attachment-processing inputs are used for the relevant request or active session. This does not delete copies already held by Discord, an external provider, or a separately enabled memory feature.
Indexed conversation memory, saved facts and profiles, preference or instruction history, agent-run records, expanded responses, saved locations, and usage records do not currently have one universal automatic deletion deadline. Some can remain until removed by a feature control or an operator. Deleting the original Discord message or removing the bot does not automatically remove all of these copies.
Reflection's default cleanup removes completed sessions after 30 days and selected run, budget, and frequency metadata after 90 days. Active observations and ideas have separate lifecycles and may remain until removed or superseded. Deployment settings can change those defaults; contact us for the schedule applicable to your data.
You can request deletion using the contact method below. We will delete data we are required to delete under applicable law and platform requirements. Any permitted retention for legal obligations or security must be limited to the necessary data and purpose. Copies held by external providers and any backups require separate handling; we will explain applicable limitations rather than promise immediate deletion from every system.
6. Your controls and privacy requests
- Use
/memoriesto inspect remembered facts and/profileto view a distilled profile when available. These commands do not provide a complete data export. - Use
/forget <text>to delete saved facts matching that text. It does not erase indexed messages, profiles, usage records, or every other stored record. - Use
/reflection_offto remove your pending sessions and associated derived reflection observations. Invoking the bot again re-enables reflection for a new session if the operator has enabled that subsystem. - Stop invoking the bot, or ask a server administrator to restrict or remove it. Nearby messages may still be processed when another person invokes it in a channel it can access.
For access, correction, deletion, export, objection, restriction, or withdrawal of consent where applicable, contact [email protected], attention [redacted]. Identify your Discord user ID and the relevant feature or server privately. We may request proportionate verification to protect your data; never send a Discord password or access token. We will respond within the deadlines applicable to your request. You may also have the right to complain to your local privacy regulator.
7. Security and international processing
MultiVAC uses application permissions, scoped data access, and bounded processing controls. No online system can guarantee complete security. If you suspect unauthorized access, contact us promptly.
The operator and providers may process information in countries other than where you live. Those countries may have different privacy laws. Where law requires transfer safeguards, the applicable safeguards must be in place for the processing concerned. Contact us for information about the providers and transfer arrangements applicable to your request.
8. Children
MultiVAC is not intended for anyone under 13 or under Discord's higher minimum age where applicable. If you believe a child below the applicable minimum age has submitted personal information, contact us so we can investigate and take appropriate action, including deletion where required.
9. Policy changes and contact
We will publish updates with a revised effective date and provide reasonable notice of material changes through the Service or its public policy location. Where required, we will obtain consent before new processing begins.
Privacy contact: [email protected]
Contact person: [redacted]
Operator: Intracoast Network Services
Operator location: North Carolina, United States
Website sign-in and server invite requests
Discord sign-in shares your Discord user ID, username and server list. We use server-management permissions to verify invite requests. The website uses a secure session cookie for sign-in and a short-lived cookie to protect the authorization flow. Sessions expire after one hour and are cleared when the service restarts. Discord access tokens remain in server memory and are never sent to the browser.
Invite requests retain the requester’s Discord ID and username, server ID and name, intended-use description, review status, notes and an access-decision audit trail. The application owner can view these records and aggregate provider usage by server and model, including token counts and estimated costs. The portal does not display chat transcripts, private memories or prompt contents. Requests and decisions are retained while needed to administer server access. Contact [email protected] to request review or deletion; access-control records may need to be retained to enforce a rejection or revocation.