Loading ToolBench
A Model Context Protocol server that exposes Mewgenics save file analysis and breeding recommendation tools through a deliberately chaotic persona. Analyzes Mewgenics save files (SQLite databases with LZ4-compressed cat data) and provides breeding optimization, team composition, and genetic simulation.
Server has 8 tools with complete input schemas and descriptions. Naming follows verb_noun pattern (analyze_, generate_, simulate_, rank_, optimize_, get_, find_, check_). Descriptions are domain-specific and adequate (100-200 chars). However, output schemas are not documented in the source code, LLMs cannot see what fields to expect from responses. Parameters lack enums where appropriate (e.g., 'goal' and 'battle_type' accept free-form strings instead of constrained enums). Error handling is not visible in the provided code. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being READ_ONLY. Parameter descriptions are present but generic; they do not explain constraints, ranges, or expected formats.
Load and analyze a Mewgenics save file. Returns summary of cats, stats, mutations, current resources, and potential breeding pairs. Use this first to understand your colony.
Analyze inbreeding coefficient and genetic diversity for a breeding pair or the entire colony.
Find all cats carrying a specific mutation. Useful for planning mutation-focused breeding programs.
Suggest an optimal breeding pair to achieve a specific goal (maximize_physical_dps, maximize_magic_dps, maximize_tank, maximize_support, pure_bloodline, collect_mutations, balanced_stats). Takes into account stat compatibility, mutations, and disorders.
Retrieve detailed information about a specific cat including all stats, mutations, abilities, disorders, and breeding history.
Generate and rank the top combat teams from your colony. Considers synergy, class diversity, role coverage, and estimated win probability for different battle types.
Output schemas not documented. LLMs cannot see what fields responses contain, forcing them to guess at downstream field names and risk broken tool chains.
Free-form string parameters lack enum constraints. 'goal' accepts arbitrary strings instead of enums (maximize_physical_dps, maximize_magic_dps, etc.). 'battle_type' similarly unconstrained. LLMs will hallucinate invalid values.
No tool annotations despite all tools being READ_ONLY. Missing readOnlyHint metadata prevents agents from understanding which tools are safe to call speculatively vs. which require confirmation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
Rank all cats in the colony by breeding potential for a specific goal. Returns top N cats with scores, stats, mutations, and disorders.
Run Monte Carlo simulation of breeding outcomes between two cats. Returns stat distributions, mutation probabilities, and disorder rates across many trials.
Parameter descriptions lack constraint details. 'trials' defaults to 1000 but no min/max bounds stated. 'top_n' defaults to 10 but no range documented. 'team_size' defaults to 3 but no validation range visible.
No error handling guidance visible in source. Tools do not document what happens on invalid save_path, missing cat_key, or out-of-range parameters. LLMs cannot self-correct on failures.