Privacy Policy

Last updated: 2026-04-26

KULVEX is built on a simple principle: your data lives on your server, not ours. The system is designed to run entirely on hardware you control. This policy explains what stays local, what leaves your machine when you opt into cloud features, and what rights you have over your data.

1. What KULVEX is

KULVEX is a self-hosted AI assistant: a server you install on your own hardware (Linux or macOS) that runs language models locally and stores all conversation history, agent configurations, memories, and tool outputs in databases on that same machine (MongoDB and SQLite). KULVEX does not phone home with telemetry, usage analytics, or chat content under normal operation.

2. Data that stays on your server (always)

The following data is generated and stored on YOUR machine. It is never transmitted to KULVEX (the company) or any third party unless you explicitly enable a feature that does so:

Chat history with all agents (channels: web, Signal, WhatsApp, Telegram, etc.)
Agent definitions, custom prompts, learned rules, style profiles
Long-term memory contents (mempalace database)
Voice recordings, image outputs, file uploads
Smart-home state, camera footage, sensor data
Mempalace embeddings, vector indexes, tool call logs
Local model weights and inference state

Uninstalling KULVEX or wiping the server removes all of this. There is no remote backup unless you configure one yourself.

3. Data sent to third parties (strictly limited)

KULVEX talks to outside services only in three cases, each of which you control:

3.1 License validation (required for paid licenses)

When you activate a paid license, KULVEX sends the following to AstroLexis (the company that issues KULVEX licenses):

Your license key
A hardware fingerprint (CPU/MAC-derived hash, not personally identifying)
The KULVEX version installed

The license server stores: the license key, the hardware fingerprint it was bound to, the email used at purchase, activation timestamp, and renewal status. Periodic re-validation may occur (default: every 60 days, with a 60-day offline grace window). No chat content, agent data, or local files are ever sent to the license server.

3.2 Payments (Stripe)

Purchases on kulvex.ai are processed by Stripe. KULVEX does not store credit card numbers, CVVs, or full billing addresses — Stripe handles all of that. KULVEX receives only: a transaction ID, the email you used at checkout, and the product/tier purchased.

3.3 External Learning (the Correction Engine — opt-in)

The most important section to understand. KULVEX has an optional feature called External Learning (also called the Correction Engine) that uses an external AI provider — xAI Grok, Anthropic Claude, or OpenAI GPT — to automatically review your agents' conversations and propose improvements to their behaviour.

This feature is OFF by default for every agent. You enable it per agent in the Agent Builder. Before the first time you turn it on, KULVEX shows a confirmation dialog that summarises this section.

When External Learning is ON for an agent, the following data leaves your server and is sent to the external provider you configured:

Excerpts of recent chat history for that agent (up to ~30 turns at a time)
The agent's system prompt and instructions
The agent's name and ID
Filter-event metadata (samples of leaked text caught by the runtime sanitiser)

The external provider then receives, processes, and (per their own retention policy) may store this data on their infrastructure. KULVEX has no control over what the provider does with the content — that's governed by their privacy policy:

If you have conversations containing sensitive information — personal identifiable data, financial records, medical details, trade secrets, third parties' personal data without their consent, or anything covered by GDPR/HIPAA-style regulations — we strongly recommend you do NOT enable External Learning for those agents. Use it only for casual or public-facing agents.

Toggling External Learning OFF stops all outbound transmission for that agent immediately. Past data already sent to the provider remains subject to their retention policy.

4. Channel integrations (Signal, WhatsApp, Telegram, etc.)

Messaging integrations route through their respective platforms. Messages traversing those platforms are subject to each platform's privacy and encryption policies — KULVEX has no control over how Signal, WhatsApp, or Telegram handle data on their servers before it reaches your KULVEX instance. Once a message arrives at your KULVEX server, it is stored locally per Section 2.

5. Cookies, telemetry, analytics

The KULVEX web dashboard uses session cookies for authentication only. There is no first-party analytics, no third-party tracking pixels, no fingerprinting, and no ad networks. The product does not phone home with usage statistics.

6. Your rights

Because KULVEX runs on your hardware, you have direct control over your data:

Access: all data is in MongoDB / SQLite on your server. You can query, export, or inspect it directly.
Deletion: uninstalling KULVEX or wiping the database removes everything. Per-conversation deletion is available in the chat UI.
Portability: agent definitions and chat logs are plain JSON / SQLite — no proprietary formats.
Opt-out of cloud features: any toggle you flipped ON can be flipped back OFF at any time.
License/account data: contact KULVEX support to delete your license record from the licensing server. Note this revokes future updates and re-activation rights.

7. Children

KULVEX is not intended for users under 16. We do not knowingly collect data from children.

8. Changes to this policy

Material changes will be reflected in the "Last updated" date at the top. For significant scope changes (new third-party recipients, new data categories transmitted, etc.), we will also surface a one-time notice in the dashboard the next time you log in.

9. Contact

Questions, concerns, or data-deletion requests: contact AstroLexis via the email associated with your license, or through the support channel listed at kulvex.ai.

Note: This is a v1 of the privacy policy. It will be revised by legal review before the v1.0 commercial release. If you spot inaccuracies or have questions about how a specific feature handles data, please reach out at the contact above.