Model Context Protocol server for TexasSolver poker solver. Provides tools to execute poker game tree analysis, load and manage poker ranges, and configure solver parameters.
TexasSolver MCP has solid foundational structure with 5 tools featuring proper JSON Schema definitions and reasonable descriptions. All tools have clear naming conventions (verb_noun pattern: run_solver, load_range, list_ranges, build_game_tree_config, list_outputs) and explicit schemas with required fields. However, there are meaningful gaps: parameter descriptions lack constraint details (ranges, formats, patterns), output schemas are not formally documented, error handling is generic without recovery guidance, and several parameters omit descriptions. The tool names follow arcade.dev conventions but descriptions could be more LLM-optimized per the 50-200 char baseline.
Generate a game tree configuration without running the solver. Useful for inspecting or editing the configuration before execution.
List previously generated solver output files.
List available preflop range files from the TexasSolver ranges directory.
Load and parse a poker range file from the TexasSolver ranges directory.
Execute the TexasSolver with custom game parameters to analyze poker game trees and generate strategies.
Output schemas not documented for any tool. Agents cannot infer what fields to expect or plan downstream operations (e.g., after listing_outputs, what field contains the filename to pass to load_range?).
Parameter descriptions lack constraint information. Examples: 'pot' and 'effective_stack' have no minimum/maximum bounds; 'range_ip' and 'range_oop' have no format specification; 'board' accepts empty string but no example format given beyond 'As,Kh,Qd'. Agents may pass invalid values without self-correction.
Error responses are generic ('Error: {result.error}') without actionable recovery guidance. Agents have no way to distinguish retryable errors (e.g., solver timeout) from user-fixable ones (e.g., invalid range syntax) from fatal failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 54 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
'bet_sizes' and 'solver_config' parameters are typed as 'object' with minimal description. No nested schema defined. Agents cannot discover valid keys or value types without trial-and-error or external documentation.
No documentation of tool dependencies or sequencing. Agents do not know if list_ranges should be called before load_range, or whether run_solver can be called directly with range strings or requires preloading.
No explicit statement of side effects. 'run_solver' description does not confirm it writes output files or modifies server state. Agents cannot determine if the tool is safe to retry or whether repeated calls with identical inputs are idempotent.