Your browser language looks different from this page. You can switch manually; we will not redirect you automatically.
Use MelaSolver GPU through one self-serve API. Preflop and presolved flop spots use published strategy data; custom flop, turn, and river spots run through our real-time solver pool. One key, stateless requests, a free tier.
pip install pokerai-bet / npm install @pokerai/client / npx @pokerai/mcp
{
"situation": "Raise",
"strategy": [
{ "action":"fold", "frequency":0.254 },
{ "action":"call", "frequency":0.106 },
{ "action":"raise", "frequency":0.64, "amount_bb":9 }
]
}
MelaSolver GPU finished in 27.75s, versus 56.47s for MelaSolver CPU and 67.22s for TexasSolver.
Scope: one M4 flop benchmark, not a global solver ranking. MacBook Pro · Apple M4 · 10-core CPU · 10-core GPU · 16 GB unified memory · 15 threads · 100 iterations · accuracy 0.5 · isomorphism on.
FullEngine V2 · 1.00x reference
2.03x slower than MelaSolver GPU
3.41x slower than MelaSolver GPU · directional estimate
TexasSolver source claimChoose stack depth, position, and action line; the complete 13×13 range and request payload update from the same data the API serves.
| Loading… |
Browse all 40 range pages → · Get your API key — the free tier covers this call.
Preflop and presolved flop spots are fast lookups, while custom flop, turn, and river spots can be submitted as queued real-time solves.
RFI, vs-raise, 3-bet and 4-bet lines across supported stack depths.
Use 5.8M+ presolved flop solutions, or submit a custom flop spot to the solver pool.
Continue from any supported spot and poll until the solver pool has a result.
Resolve final-street frequencies without shipping your own solver stack.
GTO Wizard and RangeConverter are excellent apps to study inside. Pokerai is the strategy layer itself, exposed over HTTP for trainers, bots, chart tools, review flows, and AI agents.
Serious solver infrastructure you can plug into your own product today.
A real-time GTO trainer for iOS: every hand, one thin client, no cached strategy. User action is graded against returned GTO frequencies and explained by an LLM coach.
strategy = pokerai.preflop( hole_cards="AhKh", hero="MP", actions=[("UTG","raise",3)], ) grade(hero_action, strategy) # vs GTO freq
Generate spots, grade each action, and explain frequencies after the hand.
Turn supported stack depths and positions into product-native charts.
Compare played lines against solver strategy without storing your own tree.
Let an LLM request exact strategies, then translate them into user feedback.
Presolved lookups and real-time solves are metered separately. Alipay, WeChat Pay, and crypto checkout are supported.
Builder and Pro are available with Alipay, WeChat Pay, and crypto checkout. Choose a paid plan.
Yes — 1,000 presolved lookups and 25 real-time solves per month, no credit card. Enough to build a full trainer and run every tutorial.
6-max No-Limit Hold'em, all four streets — done to solver depth, not a shallow multi-format menu. Preflop across multiple stack depths (100bb + 40bb), flop from 5.8M+ presolved solutions, turn/river solved on demand. PLO and ICM aren't covered yet.
No. Pokerai API is a hosted HTTP API for documented 6-max NLHE strategy data. Use desktop solver software for local tree construction or research control; use benchmark or gameplay APIs to evaluate agents or simulate games, not as GTO strategy output.
No. Every request is self-contained — send the spot, get the strategy. Nothing about your hands is stored.
No. Real-time assistance at real-money tables is prohibited. Pokerai API is for training, coaching, hand review, study, and research.