Your lobby will not fill from matchmaking: how co-op parties actually form
Here is the first thing that goes wrong: a new player opens the multiplayer menu expecting quickplay, waits for a lobby to fill, and nothing happens, because there is no public matchmaking and no lobby browser in Slay the Spire 2. Co-op runs 2 to 4 players, Steam friends only. The host creates the room from the multiplayer menu and sends invites through the Steam overlay, and that host has to stay online for the entire session; if the host drops, the session drops with them.
Session rules matter more than they look. Saves can be resumed, but only with the original roster: you cannot substitute a friend mid-run because someone's weekend ended, so schedule the session length honestly before you start. You cannot run multiple campaigns at once either, so the party is committed to the one climb. Ascension uses the party's lowest level and progresses separately from solo, which cuts both ways: a high-Ascension veteran carries extra risk for a friend's lower ladder, and their own solo progression stays untouched by whatever happens in the party run.
Mods deserve a paragraph of caution. All players must run identical mods in identical load order or the game desyncs, and since v0.108 modded and non-modded saves are split outright. The safest configuration for a first co-op run is mods disabled entirely on every machine. Take five minutes before the run to check, because a desync two acts deep is the most preventable way a party run dies.
Platform: co-op is PC only, with Windows, Mac, Linux and Steam Deck all counting, so mix those freely, but nobody is joining from a console.
Two content differences are worth flagging for completionists. The Merchant hidden boss fight, unlocked by throwing the Foul Potion from the Potion Courier event at a fake merchant for 300 gold, the knock-off relics and the exclusive Merchant's Rug, does not appear in co-op. If a 100 percent completion run matters to anyone in the party, do that one solo. The rest of the event pool carries over normally.
Finally, the honest prerequisite: the game's own FAQ advice is to learn solo first, and the co-op community repeats it. Co-op stacks coordination load on top of the usual decisions; a player who has never seen the Owl Magistrate's Verdict loop will not suddenly read it correctly while three other people are resolving cards at the same time. One or two solo clears, or at minimum a run deep into Act 2, is the right entry ticket. Once everyone is in the room, the host's first act inside the run should be opening settings and enabling Fast Mode, because four-player animation pacing drags badly without it.
Simultaneous turns: four players, one enemy phase, zero queue
Everyone plays at the same time. That sentence reorganizes everything about how you fight, because the solo habit of watching the enemy, thinking, then acting has no co-op equivalent: all players are resolving cards while the enemy phase approaches, and enemies attack ALL players at once. There is no queue to wait in and no teammate turn to react to.
The first consequence is that coordination has to happen before the fight, not during it. The workable routine for a party: when the fight loads, one player calls the kill order (a few seconds is enough), one player claims the debuff job, and everyone else orients energy toward damage or Block. Against Bygone Effigy the call is simple: set up powers during the sleep phase and hold Block for the wake-up turn. Against the four-body Phantasmal Gardener fight the call is kill one at a time with single big hits, because Skittish 6 grants them Block on the first hit each turn and punishes multi-hit patterns. Making that call in the first five seconds, rather than discovering the Skittish interaction after three players have already dumped AoE, is the whole difference.
The second consequence is that fixed enemy loops become party calendars. Bosses run fixed turn loops in StS2, and in co-op the whole party can block the same telegraph at the same time. Owl Magistrate: Scrutiny 16, a 4x6 peck, a defensive Ascend, then Verdict for 33 plus 4 Vulnerable; the party dumps damage during Ascend and everyone blocks 33-plus on Verdict. Mecha Knight: Charge 25, Flamethrower, Windup, Heavy Cleave 35; nobody attacks into Windup and everyone holds 40-plus Block for the Cleave. When all players respect the same loop, the effective damage a party survives is dramatically higher than four solo players each doing their own thing.
Between fights, two flows replace solo habits. Relics drop in player-count batches and are drafted one each, with ties resolved by an actual game of rock-paper-scissors. Shops are per-player: your gold, your stock, your prices. So the party never competes for shop cards but always competes for relic drops, which is a composition question with its own guide.
One more practical note: turn one is where simultaneous play wrecks the most runs. With everyone acting at once, four players each playing their own setup card can leave the enemy's first attack completely unblocked across the entire party. The convention most groups converge on is that the designated tank blocks on turn one of every fight while the damage seats develop, deviating only when the enemy's first intent is harmless. Agree on it before the first combat, because nobody has time to negotiate it mid-animation.
Team cards and passing: The Ball, Soulbound and the v0.108 set
There is no free trading in co-op. Cards do not move between players as a general mechanic; they move only through specific co-op cards, and that scarcity is what makes passing a strategy instead of a convenience.
The Ball is the flagship: a colorless 1-cost Attack that deals 10 damage, passes to a random teammate after being played, and permanently gains 15 damage every time it is passed. The strategic reading is that The Ball is party infrastructure rather than a card. You cannot aim it, the pass target is random, so the only decision you control is whether it keeps moving. The correct answer is nearly always to play it whenever you draw it: a Ball picked up in Act 1 comes back around in every subsequent fight bigger, and by Act 3 the party is passing around a 40-plus damage nuke that nobody spent a card reward slot on. The mistake is evaluating it as a mediocre 1-cost 10-damage attack and skipping it at the reward screen; in a duo it cycles between two decks, in a quad it can gain 15 damage several times per shuffle.
Soulbound is the Necrobinder's team card: it links a teammate so that your Souls enter their deck too. Souls are 0-cost tokens that draw 2 cards and then exhaust, so Soulbound effectively exports card advantage and exhaust fodder to a chosen player. The linking decision should be deliberate: link the seat whose engine converts Souls best, which in practice means the player running Haunt (whenever you play a Soul, a random enemy loses 6, or 8 upgraded, HP) or any draw-hungry engine that wants cheap cycling. Do not link a deliberately thin deck that treats every shuffled-in card as pollution; the Claw archetype, for instance, specifically suffers when extra cards enter the pile.
v0.108 added 15 dedicated co-op cards, and the patch list names nine of them: Midnight, Blaze, Blade Symphony, Concoct, Plot, Constellation, Underworld, Hibernate and One For All. Stratagem was separately unbanned in multiplayer, and Tag Team was buffed in v0.104, both signs that Mega Crit is actively tuning the multiplayer card pool. Expect the co-op card list to keep moving between now and 1.0, and re-check names before you build a strategy around one.
The general decision rule for any team card: before you take it at the reward screen, name the teammate whose deck it improves and the fight where it pays off. A co-op card that helps a hypothetical teammate is a dead card in your deck; one that feeds the actual Haunt player or the actual burst seat is a doubled payoff, because in a party a card two players benefit from is worth roughly two cards. And when a pass does happen, the receiver should plan the same turn: Souls exhaust after drawing 2, so unspent hand space wastes the transfer.
Downed and revive: why the tankiest player is the last one allowed to die
The revive rule is short and brutal: a player who falls in combat revives at 1 HP on the next floor as long as at least one teammate survives the fight. A full wipe ends the run. All of co-op survival strategy falls out of that pair of sentences.
The iron rule the community derived from it: keep the tankiest player alive. If anyone must go down, it must not be the tank, because the tank going down leaves the squishy seats exposed to the enemy phase without the party's Block anchor, which converts one downed player into a full wipe in a single turn. Practical consequence: when the team's HP runs low, funnel defensive potions to the tankiest player, not to whoever is lowest or loudest. A Block Potion (12 Block) on the tank prevents more party deaths than the same potion on a damage seat, because the tank is the one whose job is eating the 33 Verdict.
The post-revive protocol matters just as much as the prevention. A revived player comes back at 1 HP, and 1 HP means any unblocked hit downs them again immediately. Campfire Mend, the co-op rest option that heals another player for 30 percent of their Max HP, exists precisely to lift a revived ally off 1 HP. The decision rule at a campfire with a revived teammate: Mend them before anything else; an upgrade is worth less than not starting the next fight with a one-hit-down party member. Beyond that emergency, Mend competes with rest and upgrade normally: Mend the tank when they sit below roughly 40 percent and an elite or boss is the next node; otherwise upgrade key cards.
Potions deserve their coordination note: since v0.109, nearly any potion can target teammates. That turns Block Potion, Weak Potion (3 Weak) and Vulnerable Potion (3 stacks) into party resources. The rule: defensive potions flow to the tank, offensive debuff potions get played by whoever's window benefits most, Weak before the boss's big attack, Vulnerable before the party's burst. One exception is carved in stone: Fairy in a Bottle still only saves its owner, so it is personal insurance, never party insurance. Give it to whoever plays the riskiest seat, and never count it in the party revive math.
On personal safety layers: Lizard Tail revives its owner at 50 percent Max HP once per run, and Beating Remnant caps HP loss at 20 per turn. Both belong on the seat that intends to be the last one standing, which is usually the tank. The framing to internalize is that death in co-op is a resource with a price: the run pays 1 HP and one player's next fight whenever someone drops, and it pays everything when the tank drops first. Spend downs with the right player, and treat the revive-at-1-HP mechanic as a scheduled inconvenience rather than a safety net you can lean on twice per act.
Enemy HP scaling: the official wiki formula, and why the 1.75/2.5/3.25 numbers are wrong
This section exists because the co-op scaling numbers floating around the community are wrong, and planning damage checks around wrong numbers loses runs.
The official wiki formula: co-op enemy HP equals base HP multiplied by player count, then multiplied by an act coefficient of 1.1 in Act 1, 1.2 in Act 2 and Act 3, and 1.3 for the Act 3 boss. Work it through for a four-player party: Act 1 enemies sit at 4 times 1.1, which is 4.4x base HP; Act 2 and Act 3 at 4.8x; and the Act 3 boss at 5.2x. That matches the wiki's own summary that four players face roughly 4.4 to 5.2 times the solo HP bar.
The numbers you will see quoted in older threads and videos are multipliers of 1.75, 2.5 and 3.25 by player count. Those figures do not match the official wiki, and the gap is not academic. Take a duo planning around Waterfall Giant in Act 1b: base 240 HP. The myth says roughly 1.75x, so the party budgets for around 420 HP; the wiki formula says 2 times 1.1, which is 528 HP. That is more than a hundred HP of misplanning on a boss whose Steam Eruption self-destruct punishes slow kills. Use the wiki formula.
What does not scale is just as important: enemy attack damage does NOT scale with player count. The 33 Verdict hits for 33 whether one player or four faces it. This asymmetry is the whole shape of co-op difficulty: incoming damage per player stays solo-priced while the HP bar multiplies, so parties are rarely burst down but frequently timed out. If your co-op losses look like slow grinds where the boss eventually stacks out of control, Ceremonial Beast gaining 2 Strength per turn, Knowledge Demon gaining it permanently, that is the scaling formula telling you your damage did not keep pace, not bad luck.
What else scales besides HP: Plating, Artifact, Slippery and Skittish stack higher with more players, and enemy Block also scales. Two patch notes refine the picture: v0.104 shipped explicit co-op scaling nerfs to Artifact, Slippery and Skittish, and v0.108 cut the 2-player enemy Block scaling to a flat 2x. Read those together and you get the design intent: multiplayer enemies are meant to be tankier against the party's multiplied damage, without defensive auras like Artifact making debuff strategies unplayable.
And this is where the design lands: percent-based executes, meaning the Necrobinder's Doom, are prized specifically because a Doom threshold reads the enemy's scaled max HP, so the mechanic does not care how big the multiplier got. The next section turns the formula into per-act damage budgets you can actually check a deck against.
Running the numbers: how much damage per act your party must output
Here is the formula turned into numbers you can check your party against, using base HP from the bestiary and the wiki coefficients. Some land on decimals because the multiplier is not tidy; round up when planning.
Duo, Act 1 (2 players, 1.1 coefficient): Byrdonis goes from 91 to about 200, Terror Eel from 140 to 308, and the Act 1a bosses land between 380 (Vantom, base 173) and 554 (Ceremonial Beast, base 252). The Act 1b Waterfall Giant is 528. A duo needs roughly double its solo damage by the boss, which is achievable because two decks each bring a solo-worthy engine.
Trio, Act 2 (3 players, 1.2): Infested Prism, the 200 HP gatekeeper that checks whether your build came online, becomes 720. Entomancer is 522, and Knowledge Demon, widely rated the hardest boss, goes from 321 to roughly 1156. That is the act where an under-scaling trio stalls: three players whose decks peak at 30 damage per turn each need well over a dozen clean turns against the Prism while it and the rest of the act keep pressuring.
Quad, Act 3 (4 players, 1.2 on elites, 1.3 on the boss): the Knight Gang goes from nasty to structural: Magi Knight around 394, Spectral Knight around 446, Flail Knight around 485, Soul Nexus about 1123, Mecha Knight 1440. The Act 3 bosses multiply at 1.3: Test Subject #C10, a 600 HP marathon across three phases, becomes 3120, and the Queen, roughly 400 base HP before her 199 guard, is around 2080.
Practical readings from those numbers. First, flat damage stops mattering: a Fire Potion is 20 damage whether the target has 200 or 1400 HP, so percentage and compounding mechanics, shared Vulnerable, Strength that doubles with Limit Break, Poison that doubles with Catalyst, Stars banked across turns, and Doom executes, carry the acts. Second, the act coefficients are close together (1.1 to 1.3), so the real jump is player count: a trio does not need much more total damage than a duo, but each individual deck carries a smaller share, which is why role clarity matters more as parties grow. Third, enemy damage staying flat means healing and Block are solo-priced: Pantograph healing 25 HP at boss combats is worth the same to its owner in a quad as in a duo, which makes defensive relics proportionally stronger in bigger parties.
The honest budget check: if the party's combined output is not clearly more than double a solo deck's by the end of Act 1, someone is drafting like a solo player, usually by taking every card reward instead of skipping for focus. Fix it in Act 2, because Act 3 does not forgive it, and the Act 3 boss coefficient of 1.3 is applied to the largest HP pools in the game.
Doom, Poison and the tools that actually keep pace with scaled HP
Given the scaling formula, which mechanics keep pace? Ranked by how directly they answer a multiplied HP bar.
Doom first, because it breaks the category. Doom is a stacking debuff whose rule reads: at turn end, an enemy whose Doom is at or above its max HP is executed. The threshold is defined by the target's own HP pool, so when co-op scaling multiplies the pool, the requirement grows with it, but the mechanic never breaks: enough stacks still deletes the bar. The community rates the Necrobinder the only S+ co-op character on this basis. The toolkit: Scourge applies 13 (16 upgraded) and draws a card, so stacking costs no tempo; Countdown is a 1-cost power adding 6 (9 upgraded) Doom per turn, a permanent annuity in long fights; Reaper Form converts all attack damage into additional Doom, meaning the damage engine and the Doom engine become one engine; End of Days drops massive AoE Doom that immediately executes everything under threshold.
One co-op-specific timing note: v0.108 changed the end-of-turn resolution order, and Doom no longer kills first. Coordinate the execute turn: the Necrobinder should call the Doom total, because teammates dumping a finisher into a target one stack below threshold is either a wasted card or a lucky assist, depending entirely on communication.
Poison second. The build data rates Poison great in co-op, and the reasons are structural: Noxious Fumes applies 2 (3 upgraded) Poison to ALL enemies every turn, so it scales with the multi-enemy boards co-op loves, and Catalyst doubling or tripling stacks is a percentage-style payoff. Snecko Skull adds a stack to every application. Poison's cost is patience, and patience is affordable in co-op precisely because incoming damage per player stays flat; the party can afford to block and tick.
Shared Vulnerable and Weak third, covered earlier but worth ranking: one player's debuff multiplies everyone else's damage, the cheapest damage-per-energy in the mode, and it requires exactly one seat to invest.
Strength and Stars fourth. Strength compounds (Limit Break has no cap) and Whirlwind spreads it across all enemies, while the Regent banks Stars across turns, unpunished, for a 49-damage AoE Seven Stars; both are engines that out-scale flat numbers given setup turns, and co-op's flat enemy damage grants those turns. The Defect's Claw cycle kills faster than anything else per shuffle but is comparatively flat, so it leans on the shared Vulnerable layer to stay relevant late.
Attrition tools fifth: Frost Block (Glacier plus orb passives), Osty body-blocking hits aimed at players, and Beating Remnant capping loss at 20 HP per turn. These kill nothing; they buy the turns that Doom, Poison and Strength need.
The failure mode to avoid is a party of four flat-damage decks: everything dies slowly, nothing scales, and the Act 3 boss at 5.2x turns the run into a war of attrition the boss wins by default.
Relic drafting, potion funneling and per-player gold
Between fights, co-op redistributes three resource streams, and parties that manage them deliberately outperform parties that wing it.
Relics drop in player-count batches and are drafted one each, with ties broken by rock-paper-scissors. The draft is a composition decision, not a greed decision, and the working rules are these. Engine relics go to the seat whose engine they complete: Pen Nib, which doubles every 10th Attack, belongs to the multi-hit or heavy-hitting damage seat; Snecko Skull to the Poison Silent; Paper Phrog to the Vulnerable Ironclad; Akabeko's 8 Vigor to whoever opens with the biggest attack. Defensive relics go to the tank: Anchor's 10 opening Block, Sai's 7 Block per turn, Pantograph's 25 HP at boss combats, Beating Remnant's 20-per-turn loss cap. Economy and generically broken relics go wherever they compound: Black Star, which makes elites drop an additional relic, goes to whoever will fight the most elites; Ectoplasm's extra energy goes to the burst seat; Ice Cream's conserved energy goes to whoever banks resources, which in co-op is often the Regent player literally banking Stars between turns.
Gold and shops are per-player. Your gold is yours, shop stock is yours, and Membership Card's 50 percent discount applies only to its owner. Coordinate purchases verbally: two players buying removal at the same shop is fine; four players hoarding gold for the same rare is a party problem. The solo wisdom that removing a starter Strike or Defend is usually the best gold sink holds per player in co-op, and it matters more, because party damage checks punish diluted decks harder than solo ones do.
Potions have three slots each and, since v0.109, nearly any potion can target teammates, which turns the combined potion belt into party inventory. The standing rules: defensive potions funnel to the tankiest player when the team's HP runs low; Weak Potion (3 Weak) and Vulnerable Potion (3 stacks) go into boss openers, Weak before the big attack, Vulnerable before the party's burst window; Fairy in a Bottle is personal insurance for its owner only and never counts as party safety; Alchemical Coffer, which fills all 4 potion slots instantly, is at its best right before a boss in the hands of whoever will actually drink them.
Campfires add Mend to the classic rest and upgrade choice: Mend heals another player for 30 percent of their Max HP. The priority order a party should agree on before floor 1: Mend a revived ally off 1 HP immediately; Mend the tank below roughly 40 percent before an elite or boss; otherwise upgrade key cards, because HP above 1 is a resource and upgrades compound across the whole remaining run. One last note for mixed-experience parties: Ascension runs at the lowest member's level, so the veteran seat should expect to carry proportionally more of the damage budget, and should probably claim the scaling engine role for exactly that reason.
The two-player coordination checklist: run it before every fight
Everything above compresses into a routine you can actually run. Here is the two-player version; four-player is the same list with more talking.
Before the run: confirm identical mod configuration (ideally none), confirm everyone is on PC with the host online for the whole session, agree on roles at character select, and enable Fast Mode in settings.
Before each fight, five seconds of calls. One: kill order for multi-enemy fights. The classics: Phantasmal Gardener, one at a time, single big hits, because Skittish 6 grants Block on the first hit each turn. Decimillipede, focus one segment; when a segment dies while others live, it Reattaches at 25 HP, so AoE is a trap. Entomancer, burn the caster directly, because the bee board keeps growing otherwise. Kaiser Crab, focus the Rocket claw whose 31 Laser is the lethal move, kill both claws within a turn or two since the survivor gains Strength and 99 Block, and use AoE, which hits both.
Two: who debuffs. One player owns Vulnerable and Weak; the other plays damage into them. Against Artifact carriers like Mecha Knight, the owner strips the 3 Artifact with cheap applications first.
Three: block assignments on fixed loops. Owl Magistrate Verdict is 33 plus 4 Vulnerable. Vantom turn 3 is a 25-plus killer swing. Mecha Knight Heavy Cleave is 35 after a Windup nobody should attack into. Frog Knight below half HP fires a one-time 35 beetle charge, so have 35-plus Block ready on the turn you push it under. When both players know the loop, both block the same telegraphed turn and both attack the others.
Four: status-card hygiene is personal. Beckon from Soul Fysh, Burns from Mecha Knight, Infections from Phrog Parasite: each player clears their own, and each player's deck should have brought the discard or exhaust tools to do it.
Five: burst alignment. Hold finishers, Seven Stars, Catalyst, End of Days, Pen Nib windows, for the shared Vulnerable window, and let the Necrobinder call Doom totals, since v0.108 changed the end-of-turn resolution order.
Six: survival triage. Potions funnel to the tankiest player when HP runs low. If someone goes down, any survivor carries the fight: the downed player returns at 1 HP next floor, and the next campfire Mends them before anything else. Never let the tank be the first death if the party can redirect pressure, Block, potions or Osty body-blocking, anywhere else.
Run that loop every fight and co-op stops feeling chaotic. The systems underneath are the ones you already know from solo play; the only genuinely new skill is making the calls out loud, early enough for three other people to act on them.