Server has 10 tools with explicit registrations, descriptions, and input schemas. Naming follows verb_noun convention (get_*, play_*, edax_*). Descriptions are present but brief (avg ~60 chars, below the 194-char baseline). All tools have input schemas with proper types and enums. However, output schemas are undocumented, responses return TextContent + structuredContent but no formal schema declaration. Parameter descriptions are minimal. Error handling is absent, no recovery guidance, no validation errors, no retryability hints. Missing tool annotations (readOnlyHint, destructiveHint). The play_move tool modifies state but lacks confirmation or dry-run support.

• Document output schemas formally. For get_board, declare: 'Returns {player_bitboard: string (hex), opponent_bitboard: string (hex), current_player: "black"|"white", board_display: string}'. This enables LLMs to extract and chain results.
• Expand tool descriptions to 100 - 200 chars. Example: 'Get suggested moves from Edax engine. Pass n (1 - 10) to retrieve top-n candidate moves ranked by score. Use after get_board to plan next move. Returns move string, score, and search depth.'
• Add input validation and error messages. In play_move, validate coordinate format (e.g., /^[a-h][1-8]$/) and return 'Invalid move format: got "xyz". Expected coordinate like "f5". Legal moves: [list]' on failure.
• Standardize naming: rename edax_hint → get_hints, edax_get_current_player → get_current_player, edax_get_opponent_player → get_opponent_player, edax_get_last_move → get_last_move. Consistent verb_noun convention improves discoverability.
• Add tool annotations: mark get_* tools with readOnlyHint=true; mark play_move with destructiveHint=true and idempotentHint=false. This signals to agents which tools are safe to retry.