Acceptable use policy
Effective date: [EFFECTIVE DATE]
Compi does real things on your computer, with your accounts. This policy says what you may not use it for, and what Compi will not do even if you ask. It is part of our Terms of service.
What you may not use Compi for
Do not use Compi, or ask it, to:
- break the law, or help anyone else break it;
- commit fraud, scams or identity theft, or impersonate a person or organisation;
- send spam or unsolicited bulk messages, or post fake reviews;
- harass, threaten, stalk or abuse anyone;
- create, store or share child sexual abuse material, or anything that sexualises minors;
- make, acquire or use weapons capable of mass harm, or help others do so;
- create or spread malware, or attack, probe or overload systems and networks you do not own or have permission to test;
- get around security measures, paywalls, rate limits or CAPTCHAs, or create accounts in bulk;
- scrape websites at scale against their terms;
- collect, track or expose other people's personal data without a lawful reason, or otherwise violate their privacy;
- use someone else's accounts, devices or data without their permission;
- abuse Compi Cloud: get around metering or limits, use the voice gateway for anything other than app voice, share your account, or resell access to Compi Cloud.
The sites and services the agent uses have their own rules. Follow them.
What Compi will not do
These limits are built into Compi. They protect you if a model is fooled by something it reads, for example a malicious web page.
- It never pays without your approval. Every payment needs a fresh approval for that merchant and that amount, signed on your phone with Face ID or your passcode. One approval pays once. There is no "always allow" for payments.
- Some payment methods are off limits to the agent. Crypto wallets, one-tap payment sheets (Apple Pay, Google Pay), saved-card autofill and machine payment protocols such as x402 stay blocked. You do those yourself.
- It never reads your credentials. Passwords, keychain items, password manager vaults, SSH keys and API tokens are blocked. When a site needs a password, Compi fills it from your password manager straight into the verified field, and the model never sees it.
- It never solves CAPTCHAs. When a site asks, Compi hands over to you.
- You do your own Google sign-in. Your Google password and two-factor steps are always done by you.
- Risky actions wait for you. Sending messages, posting, deleting data, purchases and unknown commands wait for your tap, showing exactly what will happen.
- You can stop everything. The kill switch stops all tasks at once, from your iPhone or your computer.
- Screenshots stay private. Sensitive fields are blacked out before a screenshot leaves your computer, and screenshots are deleted after use.
- It does not set up coding agents for you. Claude Code, Codex and similar tools run on your own install and your own sign-in. Compi never installs them or signs in to them for you.
- It never asks for administrator rights. No sudo, no administrator prompts.
What we do if this policy is broken
We may refuse a request, limit features, or suspend or close the account involved, depending on how serious the problem is. Where the law requires it, we may report illegal content or activity to the authorities. If you think we made a mistake, write to hello@quetzaltech.es.
Report misuse
If you see Compi being misused, write to hello@quetzaltech.es.
Highlighted text in square brackets is a placeholder that will be filled in before launch.