Stop making the same mistake at the table.

Two minutes a day. One spot, one decision, and the solver’s answer, not a guess. In the chat you already have open.

Add to your agent

Muse · ChatGPT · Claude

Works in Muse, ChatGPT and Claude. Every verdict comes from a solver, with its source named. Nothing here is guessed.

How it works

  1. You ask for a hand

    The agent picks the concept you miss most and hands you a link.

    Deal me a hand.
    K♠K♥ under the gun, six players, 100 blinds deep. Play it here: poker.rails.so/x/…

  2. You lock one decision

    The page opens on your phone, no login. One tap locks the action and the solver grades it.

    You, under the gun. Six players, 100 blinds deep.

    K♠
    K♥

    Pot 1.5 bb · Call 1 bb

    Fold
    Call
    Raise
    All in

    Locked: Raise

  3. The answer lands back in the chat

    One tap sends the result back. The agent says what the solver said, names the concept, and deals the next drill on it.

    I raised K♠K♥ under the gun. Verdict: Good. What did I miss last time?
    Good. Solver raises 100%. Last time you folded the top of your range from early position. Here is that spot again: poker.rails.so/x/…

What the solver said

Illustrative verdicts and values. Your drill uses its own solver result.

Three verdicts, three concepts. The ruler is expected value in blinds: the filled tick is the solver’s best action, the ringed tick is the one that was taken.

Mistake

Folding here costs you 2.1 blinds.

Expected value in blinds: Fold 0.0, Call 1.9, Raise 2.1. The solver’s best action is Raise; Fold was taken. Cost: 2.1 blinds. −2.1 bb Fold 0.0 Call 1.9 Raise 2.1

Solver raises 100%, folds 0%.

Solved restricted subgame, measured exploitability.

Folding the top of your range

Close

Calling here costs you 0.3 blinds.

Expected value in blinds: Call minus 0.3, Fold 0.0. The solver’s best action is Fold; Call was taken. Cost: 0.3 blinds. −0.3 bb Call −0.3 Fold 0.0

Solver folds 72%, calls 28%.

Solved restricted subgame, measured exploitability.

Calling too wide from the small blind

Good

Nothing lost. You took the solver’s line.

Expected value in blinds: Check 0.6, Bet 2.4. The solver’s best action is Bet, and Bet was taken. Cost: 0.0 blinds. 0.0 bb Check 0.6 Bet 2.4

Solver bets 84%, checks 16%.

Solved restricted subgame, measured exploitability.

Betting every flop you raised

The concepts

Before the flop

Folding the top of your range

Defending your blinds

Calling too wide from the small blind

After you raised

Betting every flop you raised

Facing a bet

Folding too fast to a turn bet

Getting paid

Checking back the river with the best hand

Bluffing

Bluffing with a hand that can still win

Twenty-four mistakes, named the way you make them. The one you miss most is the next drill.

Connect it

Add the endpoint to any host that speaks MCP.

https://poker.rails.so/mcp

Read the docs

Muse · Directory availability pending.
ChatGPT · Settings, Connectors, Advanced, add it as a custom connector.
Claude · Settings, Connectors, Add custom connector, paste the endpoint.
Claude Code · claude mcp add poker https://poker.rails.so/mcp

Pricing

Free. Every verdict, every why.

5 drills a day

Free for now.

Pricing

Five drills a day are included. Pricing.

Questions

Is this real-money poker?

No. Play chips have no cash value and the app connects to no poker site.

Where do the verdicts come from?

A solver, precomputed; the source is named on every verdict; the agent never invents a number.

Which agents?

Muse, ChatGPT, Claude, and any host that speaks MCP.

What does the agent see?

Your drills and verdicts, nothing else; export and delete from your dashboard.

Does it tell me what to do in a live hand?

No, and it says so once, then offers a drill.

Deal me a hand.

Add to your agent

Muse · ChatGPT · Claude