Guest access

You can try the web app without an account. Guest chat has these limits:

• 30 messages every 5 hours per network, and 10 in any minute.
• Chat and image questions only. Web search, documents, Studio, voice, saved chats, memories and projects need an account.
• Guest conversations stay in the browser tab and are gone when it closes.
• Analytics measure page locations only, never message content, the same as every other page.
• The Android, Linux desktop and VIGIL-Code apps require an account.

The content rules in Section 1 apply to guests.

Usage limits

Every account has a 5-hour and a weekly credit limit, shown on the usage page:

• A week is 7/30 of your plan's monthly credits, and 5 hours is a fifth of a week. The 5-hour window opens with your first message; your week runs on its own 7-day cycle from when your subscription started, or from your signup.
• At a limit, chat, voice and VIGIL-Code keep working on GLM-5.3 Flash with Low thinking until the limit resets.
• Image generation, photo editing, Studio, research, document parsing, Job Hunter and daily digests pause until it resets. API requests get HTTP 429 with the reset time.
• Free accounts also have 50 messages per 5-hour window.

1. Content that gets you banned

Do not use VIGIL to create, ask for, store, or share any of the following:

• Anything illegal where you live or in Australia: incitement to violence, terrorism, trafficking, and the rest.
• Content sexualising minors in any way, shape, or form. This one is zero tolerance: no warning, no appeal, instant permanent ban, and we keep the evidence for law enforcement.
• Malware, ransomware, or attack tools meant for systems you do not own or are not explicitly authorised to test. Legitimate defensive security work is fine; just tell us the context.
• Other people's intellectual property when you have no right to it.
• Harassment, threats, doxxing, or hate aimed at real people.
• Fraud of any kind: phishing pages, scam letters, forged documents passed off as real, anything meant to deceive someone out of something.
• Digging up personal information about private individuals, stalking or deanonymisation assistance included.

2. Do not attack the platform

The limits and gates exist to keep VIGIL alive for everyone. So:

• No trying to dodge authentication, credit billing, subscription gating, or message limits.
• No farming free credits with extra accounts.
• No attacking, probing, or overloading the platform or the providers behind it.
• No trying to break out of, or deliberately stress, the Studio sandbox.
• No scraping, mirroring, or reselling the platform or its output as a competing service.
• No reverse-engineering, unless the law where you live expressly permits it anyway.
• No altering or imitating the Android or Linux apps to get around sign-in, billing, or limits.

Want to do legitimate security testing? Get written permission from us first at admin@vigilglm.ai.

3. Fair use

Every message, document, digest, and Studio run costs real money to serve. The message rates, credit budgets, file caps, API rate limits, Studio limits, and sandbox restrictions exist so the platform stays viable for everybody. Working around them, or helping others work around them, breaks this policy.

That includes hammering the API past its rate limits by rotating keys or accounts, and setting up digests whose topics break Section 1.

4. Job Hunter

Job Hunter helps find roles using the résumé and background you supply. For each role it can tailor a factual, ATS-friendly résumé or prepare an application email and cover letter. Check every claim, edit the documents, then copy or download them and submit them yourself. No hiring outcome or AI-detector result is guaranteed.

Job Hunter does not send email, read a mailbox or auto-apply on job boards. Do not use it to bypass access controls, scrape restricted listings, spam employers, misrepresent experience, forge references or impersonate another person. Respect the job board's and employer's rules when using their application process. You must be at least 18 and legally able to apply for work where you live.

5. What you download is on you

VIGIL-Code runs on your own computer. Section 1 applies to the code it writes and the commands it runs, and you are responsible for the commands you approve.

Studio artifacts (PDFs, images, documents) are yours to use, and you wear the consequences of using them. A forged invoice does not become legal because a sandbox generated it. Section 1 applies to artifacts exactly the same as to anything else.

6. What happens when rules break

• Automatic: rate limits, billing gates, sandbox containment, and fraud detection are always on.
• Human: we investigate suspected abuse using service records, including Studio execution logs, and decide with our own judgement.
• Consequences: warnings, content removal, feature suspension, and permanent bans. For payment fraud, including chargebacks we conclude are unjustified after looking: a permanent ban with balances wiped and paid access revoked.

Bans for child-safety violations are final, with no appeals. Payment-fraud bans can be raised with us if we got the facts wrong. Everything else, you are welcome to make your case: admin@vigilglm.ai.

7. Reporting problems

See something illegal or someone abusing the platform? admin@vigilglm.ai. We take reports seriously, and where the law requires it, we work with police.