MCP server for Lichess chess analysis with pattern detection and spaced repetition
Server has 17 tools with basic schemas and descriptions, but significant quality gaps. Most tool descriptions are 20-60 chars (below the 194-char baseline for A+ tools). Parameter descriptions are minimal or absent. No output schemas documented. Error handling guidance is missing. Tools like 'diagnose_player' and 'match_patterns' accept complex array inputs with minimal schema detail. No tool annotations (readOnlyHint/destructiveHint). Only 'persist_report' is marked WRITE; others lack risk clarity. Naming is generally verb-noun compliant, but some tools ('workspace_info', 'anonymous_session') are vague about their purpose and when to use them.
Analyze a single game with Stockfish engine
Analyze pending games in cache
Analyze a chess position with Stockfish
Create anonymous analysis session without API key
Generate cross-game coaching report with pattern analysis
Generate opening-specific coaching report
Analyze opponent pool and generate coaching insights
Generate coaching report for a single game
Descriptions are too short (20-60 chars vs. 194-char baseline). LLMs cannot determine when to select tools or distinguish between similar ones (e.g., 'diagnose_player' vs 'match_patterns' vs 'coaching_cross_game' all analyze player weaknesses but lack clarity on differences).
Complex array parameters ('games', 'analyses', 'metadata') lack schema detail. 'diagnose_player' accepts 'games' array and 'match_patterns' accepts 'analyses' array with no documented structure, type, or field requirements. LLMs cannot construct valid inputs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
Generate personalized training plan based on weaknesses
Generate cross-game weakness report for a player
Fetch games for a Lichess user
Import and analyze PGN file
Detect recurring patterns in player games
Explore opening statistics and variations
Save analysis report to file
Get player profile and statistics
Get workspace configuration and cache status
No output schemas documented. Tools return unstructured results; LLMs cannot plan downstream calls or extract required fields for chaining (e.g., 'fetch_games' returns game list but no documented structure for 'game_id' field that 'analyze_game' needs).
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Only 'persist_report' is marked WRITE; 16 other tools lack risk classification. LLMs cannot determine which tools are safe to retry or have side effects.
Ambiguous tool names and overlapping responsibilities. 'diagnose_player', 'match_patterns', 'coaching_cross_game', and 'coaching_training_plan' all analyze player weaknesses but lack clear distinctions. LLMs will struggle to choose the right tool.