Import your games and get high-depth full-game analysis from Stockfish 18 in seconds.
See it in action
We spin up as many AWS C7a CPU cores (3.7 GHz 4th generation AMD EPYC) as needed to analyze every position in parallel.
Import games from Lichess or Chess.com
No PGN copy-paste required.
Resumable analysis
We snapshot and store the Stockfish process for 48 hours. Stockfish picks up exactly where it left off.
Maia 3 + engine exploration
See what Stockfish recommends and what humans at any Elo would actually play (via Maia 3, released March 2026). Click any engine line to insert it as a variation, then analyze that branch as deep as you want.
Color-coded nested variations
Each branch depth gets its own color, so you can follow deeply nested analysis at a glance.
Single position analysis
Run Stockfish on a single position for minutes or hours.
Pricing
During the beta, you will not be charged real money. Every beta account gets 1,000 minutes of free CPU time. If you run out, just email me and I'll add more minutes.
Post-beta pricing (tentative)
Standard
1.00¢ / CPU min
Attempts to start all cores within 30 seconds for the fastest results.
Batch
0.50¢ / CPU min
Cores ramp up gradually. Slower start, more efficient overall.
Continuous
coming soon0.25¢ / CPU min
Automatically applied to single positions running 10 minutes or longer.
Pay only for the milliseconds Stockfish is actually running. No subscriptions.
Making this feasible: the technical details
TL;DR: we built a savagely efficient autoscaler.
Allowing for both short and arbitrarily long analysis is a technical challenge. Scaling up is easy, but scaling down is hard. Launching single-core servers for each position is inefficient and quickly hits rate limits. Launching multicore servers eventually leaves you with servers that are mostly idle but can't be terminated because they're executing long-running Stockfish processes.
The secret sauce is being able to migrate Stockfish processes between servers. Our autoscaler can move a running Stockfish process from one server to another in less than five seconds. Whenever it detects idle cores in our fleet, it can simply terminate a server, and any processes running on that server automatically resume on another.