Crux is played with clicks or taps; the buttons below the wall do everything the keys do.

Par —Plan 0Tries 0
Stamina
100
·
How a beat works
Every step of the plan is one beat: a static move to a touching hold (3 extra to go up), a dyno to a hold two away (20 extra), or a wait. Waiting on a ledge recovers 12 a beat, on the ground 20; a jug costs 2 to hang on, a crimp 5. Rocks fall on the beats marked red, gusts blow on the beats marked blue, and the wall always shows the hazards of the beat your next click would take. A pocket is safe from rock. A crumbling hold can't be waited on and is gone once you leave it.
Jug cheap to use, cheap to hold Crimp costs 8 to reach, 5 a beat to hold Ledge rest here Pocket rock can't reach you Crumble one use, no waiting Anchor the top
Replay
Every session on a wall is recorded as inputs plus the wall's seed, so it replays exactly, planning included. Download a file or Copy a pasteable string to share it; Load or Paste to watch one.
Bests on this wall, in this browser
Fewest beats—
Summited first try—

Why it's built this way

The first Crux was a real-time climber with a stamina bar, and a fair test showed what it was: a platformer. Delay every input in a good run by a quarter of a second and it still summits; delay them by one second and it falls. The plan mattered less than the fingers.

This one has no fingers. The wall is a grid of holds, time is a beat, and the route is built before anyone moves. Every step of the plan is one beat: a static move to a touching hold, a dyno to a hold two away, or a wait. Each has a stamina price that is printed on the page and computed as you plan, so the panel always says where you’d be, when, and with how much left. The hazards run on the same clock: each rock chute drops on a published cadence, each wind band gusts on its own, and the wall always shows you the hazards of the beat your next click would take. A pocket is safe from rock. A crumbling hold can’t be waited on and is gone once you leave it. Press Climb and the climber follows the plan exactly; if the plan falls, you see where and why, and the route is kept up to that point so you can trim it and try again.

Par is the fewest beats any plan can reach the anchor in; a solver computes it for every wall, including the random ones. The six set walls add one idea at a time (stamina, then a chute, then wind across a gap only a dyno crosses, then crumbling shortcuts and long runs between rests), and the bests are kept per wall.

The slow way. Static moves only, a rest on every ledge, a beat or two in a pocket while the rock goes by. It costs beats and it never falls.

The fast way. Dynos cover two holds in a beat for twenty stamina, so a plan built around them arrives with the tank nearly empty and no room to wait. The wall gates it: a gust beat blows a dyno off, a rock beat clears the column, and a crimp costs five a beat just to hang on while you wait for either to pass. The fastest plan on most walls is a mix, and finding it is the puzzle.

The delay test that sank the first version passes trivially here: a plan doesn’t care when it was clicked.