Acceptable Use Policy

Why this exists

Two reasons. The obvious one is that we do not want to host harm. The less obvious one is that we run on model access from Anthropic, OpenAI and Google, and their rules apply to what passes through us. If you break their policies, our access is at risk, and so is every other customer's.

Do not

• Use the service for anything illegal where you are, or where we are.
• Generate content that sexualises minors, incites violence, or is designed to harass a specific person. These end an account immediately and without warning.
• Try to build weapons, malware, or tools whose purpose is to cause serious harm. These end an account immediately and without warning.
• Deliberately produce content designed to deceive people about who wrote it or where it came from, including impersonating a real person or organisation.
• Attack the service: probing for vulnerabilities without permission, overwhelming it, or attempting to reach another workspace's data.
• Circumvent credit, rate limits or access controls, or resell raw model access as if it were your own.
• Upload personal data you have no lawful basis to process, particularly special-category data under Article 9 GDPR.

Automated use

Running Coldcache from a script or a scheduled job is expected and encouraged — reproducibility is the point of the product. Load that degrades the service for others is not, and we will contact you before doing anything about it.

What happens if you break this

For most breaches we contact you and ask you to stop. For serious ones — the two marked above, or an active attack — we suspend first and talk afterwards.

Suspension stops runs. It does not delete anything: your agents, versions and history stay intact while we sort it out.

Report abuse to TO BE COMPLETED — general contact email.