Sokoban AI Solver: Automatic solver for Sokoban puzzle games.

What it does

Sokoban AI Solver is a browser-based tool that plays Sokoban puzzles and finds the optimal solution automatically. Sokoban is a 1980s warehouse puzzle where the player pushes boxes onto goal squares. The AI returns the provably fewest-move solution for each puzzle, running entirely in the browser using a JavaScript port of a C++ engine.

Who it is for

This tool appeals to Sokoban enthusiasts who want to see how puzzles are optimally solved. It is designed for people curious about puzzle-solving algorithms or players stuck on particular boards. The interface allows manual play with keyboard controls or watching the AI solve the puzzle at different speeds.

Pricing

The site does not show prices.

How it stands out

The solver uses several algorithmic optimizations to achieve optimal solutions efficiently. It applies move-optimal macro-push A* search, which treats box pushes rather than individual keeper steps as the unit of search, reducing the problem space significantly. It compacts game state into compact bitmask representations, fitting millions of states into tens of megabytes. The system uses a bucket queue and open-addressed hash table for the A* frontier and visited set, keeping memory allocation minimal and cache-friendly. Deadlock pruning eliminates provably unsolvable positions early using static dead-square tables and wall-aware push-distance heuristics.

Most puzzles in the collection are solved to proven optimality in milliseconds. One exceptionally difficult puzzle (board 15) requires too much computation for browser execution, so its optimal solution was precomputed offline using parallel processing and is simply displayed when selected.

What a founder should check

A founder considering a similar project should investigate: First, the existing competitive landscape for Sokoban solvers and similar puzzle-solving tools, including any online competitors or downloadable applications offering comparable functionality. Second, the licensing and intellectual property status of Sokoban and whether puzzle collections used are original or derived from existing sources, which could create switching costs or legal constraints. Third, whether monetization is viable for a niche audience; the current implementation shows no visible pricing model, suggesting this may be offered as a free utility rather than a revenue-generating product.

Thinking of building something like this?

Every launch here is a competitor to somebody's idea. If yours is close, check it against the market before you build: the Full Check names the rivals, the prices and the gaps.