
R.E.P.O.
An online co-op horror game with up to 6 players. Locate valuable, fully physics-based objects and handle them with care as you retrieve and extract t
Launched on February 26th, 2025

An online co-op horror game with up to 6 players. Locate valuable, fully physics-based objects and handle them with care as you retrieve and extract t
Prototype sample data: the rating and review count come from the written sample reviews below, and the follower count is derived from the game's own row so it stays put between reloads. None of it is real user data.
An online co-op horror game with up to 6 players. Locate valuable, fully physics-based objects and handle them with care as you retrieve and extract to satisfy your creator's desires.
The original listing, then every content update R.E.P.O. has shipped since.

An online co-op horror game with up to 6 players. Locate valuable, fully physics-based objects and handle them with care as you retrieve and extract t
Launched on February 26th, 2025
Update Feb 20, 2026
Update titles, dates and links are real Steam announcements, captured once and stored with the catalog; routine hotfixes are filtered out. The vote count beside each one is the prototype's own counter, not anything Steam reports. Every game's updates →
The developer's own words about R.E.P.O. — the one part of this page that is not derived from the catalog.
No pitch from the developer yet
Everything above is derived from the catalog and the storefront. If you made R.E.P.O., this is where you get to say what it is in your own words.
Based on 38 reviews
Inert in this prototype — nothing you write would be stored.
Reviewers are almost unanimous that the physics handling is the joke and the monsters are the setup — the highest-rated moments in nearly every review involve a doorway, an expensive object and two people giving contradictory instructions. Proximity chat and the six-player cap get named constantly as the reason it works better than smaller-lobby horror. The consistent complaint is netcode: non-host players report rubber-banding the moment anyone picks up something heavy, and a host with a mediocre connection ruins the run for everyone. A secondary gripe is that the level pool starts repeating once a group has thirty or forty runs in, though most reviewers say they had already got their money back by then.
Summarized for this prototype from the sample reviews below.
1 review
The object weight is the whole game and it is tuned beautifully. Two of us spent four minutes trying to rotate a grand piano through a door frame that was visibly too narrow, and that four minutes was funnier than anything a writer could have scripted. Proximity chat means the panic arrives at different volumes from different rooms, so you always know roughly how badly everyone else is doing without anyone having to say it.
Non-host players get noticeably worse physics. When I host, carts behave; when my friend with the older router hosts, everyone else rubber-bands the second a heavy item leaves the ground. It also badly needs a way to mark an item as not worth the trip, because half of every argument is four people relitigating whether the statue is extractable.
Lethal Company is scarier and this is funnier, and after about twenty hours funnier wins for our group. If you want dread that holds all session, stay with Lethal. If you want a monster that mostly exists to make carrying a vase harder, this is the better buy.
1 review
It survives a group with wildly mixed skill, which almost nothing else in this genre does. Our least confident player has never once been a liability, because the job is carrying a thing slowly and not bumping into anyone, and being bad at that is the entertainment rather than a problem. Nobody has to learn a map or a loadout to be useful on their first night.
Extraction quotas ramp harder than the group's competence does, so there is a wall around the middle where a casual four-player night suddenly requires actual coordination. That is where our newer friends drop off. I would also like the cart collision to stop treating a doorway as a personal insult.
Compared with Content Warning it asks less of you and gives back more per session — no camera to babysit, no upload to optimise. Content Warning is the better one-off night, this is the one that survives being played every week.
1 review
Six-player support is doing more work than people give it credit for. Most co-op horror caps at four, which means somebody in a friend group of six is always sitting out. Being able to take everyone means the game gets picked on nights when better games do not, and once you are in, the split into two teams of three produces its own comedy without any of it being designed.
Desync after the last couple of patches is genuinely bad in a full lobby. We verified files, dropped mods, swapped hosts, and it still happens on heavy pickups with six people connected. Also the audio ducking on proximity chat gets confused when three people shout at once, which is exactly when you most want to hear them.
Against Murky Divers, which is the closest thing tonally, this has much better moment-to-moment physics and much worse stability. Murky Divers holds a lobby together more reliably; nothing in it is as funny as a friend dropping a chandelier down a staircase.
1 review
At this price you can buy it for the friend who is not sure, and that is the actual feature. We got four copies for less than one dinner and had a full evening out of it the same night, with no tutorial and no session where anyone had to catch up on progression they missed.
The scare palette is thin. After a few weeks you know every sound cue and the horror half stops landing entirely, leaving a physics game with a monster theme. Run variety plateaus faster than the difficulty does, and there is no way to nudge the generator toward layouts you have not seen.
Phasmophobia is three times the price and asks you to learn an evidence system before it becomes good. This is good in the first ten seconds and stops improving after about ten hours. For a group that plays once a fortnight, that trade is obviously correct.
Every review on this page is sample content written for this prototype. No real players were surveyed and the handles are invented. Helpful and Report record nothing; Share is real and links straight to the review.