Invisible, Inc. review

written 2026-05-14

Invisible, Inc. is a turn-based, tactical roguelike about stealth that I played in 2026-05. Overall it was pretty fun, but I felt it's a bit too easy. I played 6 runs and only lost 1.

Core mechanics

Invisible, Inc. is a game of positioning multiple units. You usually have a team of 2-4 agents, and each has a certain number of action points per turn. You can deal with guards by moving around their vision, making sound or visual distractions to lure them somewhere else, and/or attacking them.

Each mission has a main objective, like stealing an item from a guarded room, a bunch of small optional objectives like safes you can loot for money, and then an exit you have to find and get all your agents to.

A big aspect is exploration: you don't start with a map of the building, only where your agents have seen, and have to explore to find the objectives and exit. Sometimes you find both immediately, sometimes they're the last rooms you find. This can introduce a big luck factor if a mission's going bad, you need to escape and haven't found the exit, because you might have to guess. But that doesn't happen often because usually you'll be able to explore the whole place.

Despite being a stealth game, I like how gradual the consequences of being seen are. Instead of instantly raising some drastic alarm, being seen by a guard or camera generally just increases the alarm tracker by 1, which deploys new security measures every 5 points (like turning on more cameras or spawning more guards). And of course the guard who sees you will attack, and shout to alert other nearby guards. But far away guards don't get notified.

Despite the mild consequences for being seen, I think the game keeps its stealth theme well, since you almost always want to avoid it anyway, and open battles with guards are very dangerous and unsustainable. Also, the alarm tracker increases by 1 each turn, so you have a soft time limit where it gradually becomes more dangerous to stay in the building. I like how widely this lets you choose level of risk/reward: ideally you want to stay until you've found and looted every safe in the building, but often it'll be too dangerous. If things get bad and you've found the exit, you can even evac your agents without completing the main objective.

I like how there's 3 different ways to deal with guards that are very well balanced; it's very hard to decide. You can sneak past them, KO them non-lethally, or kill them.

Running past is good because it leaves no consequences (it doesn't raise the alarm or alert anyone), but it also leaves the guard in your way, limiting your safe movement in the area, and you miss the opportunity to steal from them. Guards usually carry a few credits, sometimes a key for certain optional doors, and rarely other items. But if you have an agent with an upgraded "anarchy" skill (funny name for a skill, I know), then you can pickpocket them while they're awake.

KOing guards non-lethally is usually easiest since most agents start with a basic melee weapon that can KO a guard for a few turns. But once they wake, that guard will know there are intruders in the building and start searching instead of just patrolling back and forth, which makes them much more dangerous.

Killing guards removes them from play fully, but they all have heart monitors that increase the alarm tracker by 2 if they die. Also, lethal weapons usually have very limited ammo.

Another big aspect is the hacking system. Your team has an AI called Incognita that can spend PWR to hack devices including security cameras, safes, and even enemy drones. Agents hijacking consoles gives you more PWR. Rationing PWR and prioritizing devices to hack adds a lot of depth, since you usually can't afford to hack everything in the building, and you often have to pick based on incomplete information about the building layout or how much PWR you'll have in the future.

Journey map

Inbetween missions, you can spend credits on upgrading your agents' stats and select the next mission. And unlike the equivalents in most other roguelikes I've played, I think the different types of missions are decently balanced. There are the following types:

Executive terminals - gives more options for future missions, also lets you pick a specific type to be included
Chief financial suite - gives a vault access card and credits
Vault - gives tons of credits, but you need to bring a vault access card to get the most of it
Detention center - gives a new agent to your team, but they come with no equipment
Security dispatch - gives a random valuable item, usually a weapon
Nanofab vestibule - lets you spend credits for a wide selection of items
Cybernetics lab - gives auguments (permanent upgrades that don't take inventory space) for your agents
Server farm - lets you buy new programs for Incognita to use

These are all useful reasonably often. In particular, detention center might sound overpowered, and I do think it's always worth getting a full team of 4 before the final mission, but not always my top priority. I've had situations where I felt my team needed weapons more desperately than another empty-handed agent, so I went to a security dispatch first.

Another interesting aspect is distance. The journey map is an actual map of the world with possible missions distributed across it, and instead of a fixed number of missions, the final mission happens after 72 hours. The number of hours a mission takes depends on how far it is from your previous mission.

Run configuration

There's a lot of options for configuring your run. Besides the preset difficulty options, you can adjust a bunch of settings independently, and you can also choose 2 of several agents to start with, and 2 of several programs for Incognita. All the content is cool, but I think the Incognita programs are especially cool because they change your playstyle so much. For example, the default hacking program costs 2 PWR to break 1 layer of firewall - good design, starting players with the simplest option. But another one "infects" a device making it lose 1 level of firewall per turn, and costs the number of such infections currently running. This can be way more PWR efficient, but means you can't quickly hack through many layers of firewall on a single device (you can't stack infections, and can't start with both these programs since you need one program that generates PWR).

Bad communication

The big problem is that the game is horrible about communicating its rules. Even after 30 hours, I constantly got screwed over by unexpected behavior of mechanics I thought I finally understood. It even still happens sometimes after 50 hours.

I'll list a few examples of mechanics that are badly explained:

What actions take action points: only moving and peeking. You'd think since they're called "*action* points", not "movement points", that most actions would take one. But none of these things do: opening a door, looting a safe, hijacking a console, attacking. Most of things seem like they should take longer than peeking around a corner.

Cover: there are 3 kinds of cover that work differently, and the tutorial doesn't distinguish them at all. It just introduces "cover", teaching that all cover works the same. How they work is also very counterintuitive. You don't have to be on the opposite side of cover to be hidden: if, for example, you are 6 tiles in front of a guard, 1 tile aside, and there's soft cover 6 tiles away but 0 tiles aside, it looks like you're totally visible but you're hidden. But you *do* have to be adjacent to soft cover; if you're directly behind it but 2 tiles away, it's useless. These rules make sense once you understand them, but they're so counterintuitive that they need to be taught, and aren't. It took 30 hours of gameplay and wiki reading for me to fully understand them.

https://iiwiki.werp.site/cover

What tiles do guards see when they peek: this is well explained by the wiki. Just look how complex it is; none of it is explained in the tutorial.

https://iiwiki.werp.site/peeking#peeking_by_guards

Breaching doors: sometimes, you can lure a guard through a door with an agent waiting in ambush right behind it. Other times, the guard will "breach" the door instead of opening it, knocking out your agent standing behind it. The game never explains that this can happen or when. I looked it up, found misinformation, got screwed over by it a few times, and now I *think* I understand when it happens but I'm not totally sure. I think it's when the guard is on "alert" status and their investigation target is the tile directly behind the door.

What triggers overwatching guards to shoot: so, when a guard sees an agent, they yell something like "stop!" and go into "overwatch" mode where they point a gun and get a target icon over their head. If the agent moves anywhere that isn't immediately to cover, the guard shoots. Or the guard shoots at the start of their next turn. Or the guard shoots if you take certain other actions, like opening a door. The option to open a door has a tooltip saying "the guard will react to this action", which is horribly vague, it should clearly say "you will be shot", but it's also bad because it makes you think options that *don't* have such a tooltip *won't* get you shot. And you know what doesn't have such a tooltip, but does get you shot? Peeking! That's right, just looking around gets you shot. Oh, but *hacking a console* or *picking up a gun from the ground* doesn't. Also the guard instantly shoots if *another* agent enters their primary vision.

Oh yeah, speaking of that "immediately to cover" escape, that's also mistaught in the tutorial. The tutorial clearly says you'll get shot unless you move only 1 tile to cover, but you can actually move *diagonally* into cover, and diagonal moves are considered 1.5 tiles long in this game. Similarly, some things have tooltips that claim their range is "N tiles" when it's actually "< N + 1".

Guards retain overwatch status: if you move out of overwatch, the guard says something like "dammit they got away!" and loses the target icon. But they're actually still in overwatch! If any agent moves into their line of sight in the same turn, they shoot instantly.

There's an enemy type with a tooltip saying it has 180 degree vision, but it's actually 135. There's another enemy type with the same tootlip that actually does have 180.

Emcumberment double-activates on pickup: when an agent carries too many items, they get "encumbered" lowering their AP by 1 for each extra item. So if you're carrying 2 extra items, you lose 2 AP at the start of your turn. But if you *pick up* 2 extra items, you lose 3 AP for the current turn! How can this be? I think it's because any change in your inventory reapplies the encumberment penalty.

Guards can shoot multiple times in one turn: your agents cannot, even if they have multiple weapons, even if they have a weapon whose tooltip says "infinite ammo but can only be used on one turn per mission", which makes it natural to assume guards can only shoot one agent per turn.

Subdermal tools misleading tooltip: Subdermal Tools is an augment that lets an agent disable enemy devices. On one mission, there are robotic-looking guards, and if this agent is next to them, the button to use Subdermal Tools will say "disable [guard name]". But it does no such thing. This is actually an action available on any guard, and all it does is disable their heart monitor that raises the alarm if they die.

The game does mitigate this problem with the rewind mechanic: you can rewind to the beginning of previous turn a certain number of times per mission. But it's still frustrating when needing to rewind because of rule confusion eats into your ability to rewind actual mistakes. And you only get 1 rewind per mission on expert mode, which the game presents as its "normal" difficulty (despite the name).