MCP server that lets AI clients build in a Minecraft Paper server through the Ashlar plugin
Ashlar provides 9 tools with consistent naming (mc_* prefix) and documented descriptions. All tools have input schemas with type declarations. However, parameter descriptions are minimal (typically 1-2 words), output schemas are not documented, and error handling lacks recovery guidance. Tool names follow verb_noun pattern well (mc_status, mc_build, mc_inspect), but descriptions average ~80 chars, below the 194-char baseline for A+ tools. No parameter constraints (enums, ranges) are visible. Schemas show types but lack depth (e.g., 'from' and 'to' arrays lack item type specs or coordinate format documentation).
Executes batch fill operations to build structures in the world
Executes a Minecraft command on the server
Inspects a rectangular region and returns block composition and sign data
Lists online players with their positions and game modes
Renders a visual representation of a region with multiple view options (top, facade, slice, heightmap)
Restores a previously created snapshot to undo building operations
Parameter descriptions are minimal (1-2 words). 'from' and 'to' lack format guidance (e.g., '[x, z]' vs '[x, y, z]'). LLMs cannot infer coordinate system or dimensionality.
Output schemas are not documented. Tools return structured data (heightmap, block composition, player positions) but LLMs have no schema to plan downstream calls or extract fields.
Array parameters lack item type specs. 'ops' in mc_build is 'array' but no description of operation object structure (required fields, valid types). 'from'/'to' arrays lack item count or type hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Creates a snapshot of a region before building, enabling undo functionality
Returns server status including version, player count, queue length, and uptime
Surveys a rectangular region and returns heightmap data with optional visualization
No error handling guidance. Tools like mc_build (DESTRUCTIVE) and mc_command (DESTRUCTIVE) lack recovery hints. If a build fails, LLM has no guidance on retry, undo, or next steps.
Destructive tools (mc_build, mc_command) lack confirmation or dry-run support. Agents can execute irreversible operations without safeguards. mc_snapshot/mc_restore exist but no explicit undo workflow documented.