Your orientation begins now.
An adult desktop app. You clock in, get assigned shifts, and work them under an AI manager that watches what you do and has opinions about it.
An integrated browser opens PMV Haven inside the app. You browse the real site, the supervisor watches your choices and reacts. More sources scheduled.
Runs locally. No accounts. No cloud. No one watching except the one already assigned to you.
A local LLM sits on your machine and plays your manager. It sees what you watch, has things to say about it, and gets less polite the longer you work here. No servers involved. Just you and it.
An integrated browser opens PMV Haven inside the app. You browse the real site, the supervisor watches your choices and reacts in real time. Assignments filtered to your tags and rank. Not a fixed library. More sources scheduled.
No accounts. No cloud. No tracking. Your data never leaves your machine. What happens during a Shift stays between you and the system.
Pick your track at onboarding. Sissification, Femdom, Raceplay, Gooner, Full Corruption. Each shapes which content unlocks at which rank, and how the supervisor talks to you.
Ranks. Credits. Compliance Score. Corruption Index. Four systems that feed into each other. The more you comply, the deeper it goes. Cooperation is... rewarded.
Timed Shifts. Release Windows. Lockout Periods. The schedule is not a suggestion.
| Requirement | Windows | macOS | Linux |
|---|---|---|---|
| Min OS | Windows 10 (x64) | macOS 11 (Big Sur) | glibc 2.31+ |
| Architecture | x64 | Apple Silicon (arm64) | x86_64 |
| Min RAM | 8 GB | 8 GB | 8 GB |
| Disk space | 5 GB free | 5 GB free | 5 GB free |
| Runtime | included | included | AppImage runtime |
Internet connection required. A discrete GPU is optional but accelerates supervisor response time. The local LLM (Ollama) installs automatically on first launch.
The application is not code-signed. Your operating system will request confirmation before first launch.
SmartScreen may display "Windows protected your PC".
The app isn't notarized, so macOS blocks the first launch. Move it to Applications, then approve it once.
Updating from an earlier version? macOS can’t auto-update an unsigned app, so the in-app updater will report a code-signature error. Download the latest .dmg above, quit Compliance, drag the new build into Applications and choose Replace, then run the command below (the fresh download is quarantined again).
Still blocked, or macOS says the app is “damaged”? Clear the download flag from Terminal, then open it normally:
xattr -dr com.apple.quarantine /Applications/Compliance.app
Make the AppImage executable before launching:
chmod +x Compliance-*.AppImage
Procedure unsuccessful? .
Compliance started as a tool for myself. Every time I opened a tube site to enjoy myself, I'd spend more time picking videos than watching them. Five tabs open, decision fatigue, the actual watching only starting once I had enough queued up to chain through. I wanted a feed instead. Something that would decide for me, flexible enough to follow how my preferences shifted over time, because they do shift.
The "supervisor" is that feed. The fiction around it is what makes the not-choosing the point. Letting someone else decide what you watch is, in itself, a kink. Exposure shapes desire. The more you consume, the more your taste expands. With local AI, the handoff comes without human stakes. No one to keep happy, no escalating demands, no exposure of your private life. Just you and the system, on your machine.
So it isn't really a game. It's a new way of consuming porn, with gamification stacked on top to keep things from going stale. No ending. Take tasks, complete them, score moves, taste expands. That's the loop.
This is a beta. I'm publishing it to find out who else sees it this way, and what they want me to build next. The feedback button is not decorative.
RogerKink6
Connected device control during shifts. The supervisor operates compatible Lovense hardware in sync with assigned content.
Lockout periods during shifts. Release windows granted on compliance streaks. Bring your own cage.
Optional preset for Beta-tier subjects. The workstation applies real-time nudity filtering across assigned content. The body remains denied, the assignment continues. Discipline for personnel below the authorization threshold for unfiltered access.
New approved connectors beyond PMV Haven. Format and source coverage scales with the supervisor's profiling needs.
Full workstation rebrand paired with a more visible supervisor across the interface. Currently the AI lives in emails, quizzes, and task reactions. The redesign will surface it directly into the workspace. Happy to share concepts on Patreon and just as happy to hear yours.
Community channel. Additional language support. Expanded content taxonomy. Triggered by beta cohort demand.
All future protocols are subject to beta-cohort feedback. Sequence may shift based on what testers actually want.
An adult desktop application. A local AI plays your corporate manager. It assigns Shifts, monitors what you watch through the integrated browser, and adjusts your rank based on your compliance. Fictional dystopia, real browsing of PMV Haven inside the app. 18+ only.
It builds a behavioral profile from your calibration-quiz answers and adjusts continuously based on what you select during Shifts. It watches the secondary tags on the videos you pick: if your choices align with your selected orientation's conditioning arc, the system advances your tag privileges. It also escalates its tone as your Trust and Corruption indices rise. Dialog is generated by the local LLM at runtime, not pulled from a script bank.
A modern desktop (Windows 10+, macOS 11+ on Apple Silicon, or Linux x86_64), 8 GB RAM, 5 GB free disk space, and an internet connection. See Workstation Requirements above for the full list. The local LLM installs itself on first launch.
Steam doesn't authorize this category of content. Direct purchase routes require adult-payment infrastructure with KYC requirements that most processors avoid. Patreon resolves both. Cancellation is one click from your Patreon account, no recurring-billing escape rooms. Patreon handles billing end to end, no card details ever reach the developer.
Your viewing history and supervisor interactions are stored locally on your machine. No app accounts, no cloud sync, no analytics, no ad SDKs, no telemetry. The app still uses the network for specific functions: updates, optional Patreon verification, feedback you choose to send, Ollama setup and model downloads, the pages you open in the integrated browser, and any optional device endpoint you configure. See § 08.10 for the breakdown. The app is not code-signed, so your OS will warn before launch — see the Installation Protocol above, and § 08.13 if your antivirus flags the installer outright.
Only in bursts. The supervisor generates text when the game asks for it — task assignments, audits, reviews — and is idle between those moments. The model is automatically unloaded from memory after a few idle minutes (standard Ollama behavior), while the small Ollama background service stays running, exactly as if you had installed Ollama yourself.
A task is one assigned video watched to completion inside the app — as long or as short as the videos you pick. A shift chains several tasks and closes with a Performance Review, and you can stop between tasks. The free tier covers your first 8 completed assignments.
The current video page only: its title, its tags, and how much of the video you actually watched. That is what the supervisor profiles you on. There is no browsing-history collection — the integrated browser exists solely to reach the supported sources, and everything it reports stays on your machine.
No. The app contains no screen-capture or recording code, and crash reports are stored locally — never uploaded automatically. The only record is the local log file, which leaves your device only if you attach it yourself when sending feedback.
App-service traffic is limited to update checks against our download host; Patreon verification, only if you connect your account; feedback you explicitly send (version and OS, plus the last ~50 log lines only if you tick the box); and Ollama installer or model downloads during setup. User-directed traffic goes straight to the pages you open in the integrated browser and, if enabled, to the device endpoint you configure — usually hardware on your own network. Browsing is not proxied through us, and the app sends no analytics, ad-SDK data, or telemetry.
The app monitors each integrated source and tells you when the assigned one stops responding. It then offers a prompt letting you switch to another source to continue your shift — the switch is yours to confirm, it is not automatic. If every source is unreachable (for example under region-level restrictions), assignments can't load until one comes back.
In the app: Settings → Feedback & Support — app version and OS are attached automatically, the last 50 log lines only if you opt in. On this site: the Send Feedback button below. Prefer to dig yourself? Logs live at %APPDATA%\Compliance\logs\ on Windows, ~/Library/Logs/Compliance/ on macOS, ~/.config/Compliance/logs/ on Linux.
Read the detection name before anything else — it tells you what was actually judged. Windows Defender has reported Trojan:Win32/MalUri.A!cl on our Windows installer. That is a URL-reputation verdict: MalUri means the download address was rated, and the !cl suffix means the call came from Defender's cloud service rather than from inspecting the file. Nothing was found inside the program. The alarming wording Defender shows alongside it — "executes commands from an attacker" — is boilerplate attached to the whole threat category, not an observation about this app.
Why our address was rated poorly: our installers were served from a shared *.r2.dev hostname, a Cloudflare development domain used by thousands of unrelated projects. Reputation on a shared host is partly collective, so an unsigned installer fetched from it inherits everyone else's history. We have moved downloads onto our own domain to fix the cause, and we submit these detections to Microsoft for review. The app is also not code-signed, which leaves no publisher reputation to offset the rating.
Do not disable your antivirus, and do not add a blanket folder exclusion. Verify instead of taking our word for it: upload the installer to virustotal.com. A real threat lights up across most of the 70-odd engines; a reputation false positive shows one or two. If Defender already removed the file, you can restore it from Windows Security → Protection history. Send us the VirusTotal link via § 09 below and we will confirm the file hash against what we published — and if you would rather just wait until the rating clears, that is an entirely sensible call.
One case where the warning is real: a copy that did not come from this site. If you downloaded Compliance from a mirror, an aggregator or a torrent, a repackaged installer genuinely can carry something extra, and your antivirus is doing its job. Delete it and download again from here.
Something broken? Something weird? Submit a report. Each entry is reviewed.
Internal Memorandum — Classification: Restricted
This application contains explicit adult content (18+) pulled from real websites. The corporate framing is fiction. The content is real. You are a consenting adult or you close this page. All exits are... available.