Deterministic Minecraft build architect and ChatGPT MCP App
Blockwright has 39 tools with consistent naming (verb_noun pattern) and basic descriptions, but critical gaps undermine quality. All tools lack input parameter descriptions, parameters are defined with type and description fields in schema, but descriptions are generic placeholders like 'Complete build input specification' or 'Build record to audit' that do not explain what values are valid, what format is expected, or when to use the tool vs. similar ones. No output schemas are documented anywhere. Parameter types are present (object, string, number) but lack constraints (enums, min/max, patterns). Error handling is absent, no guidance on retryability, user-fixable vs. fatal errors, or recovery steps. Tools like export_build_* are nearly identical, violating single-responsibility. The server exposes file system operations (discover_worlds, install_world_edit_schematic) without visible permission gates or audit logging. STDIO transport caps protocol readiness at 50.
Analyze how a build fits with terrain in a world region.
Perform comprehensive audit of a compiled build for quality and compliance.
Check for available Java runtime updates.
Compile a normalized build input into a deterministic placement record with audit metadata.
Confirm and finalize a terrain fit preview.
Continue an interactive palette design session.
Create a Bedrock edition mcpack from a build.
Parameter descriptions are generic placeholders (e.g., 'Complete build input specification', 'Build record to audit') that do not explain valid values, formats, constraints, or when to use the tool. LLMs cannot infer parameter semantics from these descriptions.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2025-06-18+ | v2 |
Create a delivery bundle with materials list and build instructions.
Generate a material list from a compiled build.
Delete a saved material palette.
Discover local hardware capabilities and specifications.
Discover available local LLM model providers.
Discover available Minecraft worlds on the local system.
Perform preflight estimation of build feasibility and resource requirements.
Export a compiled build as a blueprint file.
Export a compiled build as CSV data.
Export a compiled build as JSON data.
Export a compiled build as a litematic file.
Export a compiled build as mcfunction commands.
Export a compiled build as a schematic file.
Generate multiple candidate builds from a normalized input for comparison and selection.
Get a style profile by identifier.
Import a litematic file and convert to build record.
Import a schematic file and convert to build record.
Install a schematic into a Minecraft world using WorldEdit.
List available Java edition registries.
List available material palettes.
List all available style profiles.
Load a saved material palette by name.
Normalize and validate a build input specification, returning an architectural plan.
Read a specific Java edition registry.
Get metadata about available registries.
Rename an existing material palette.
Revise a build within a selected world region.
Save a material palette with a given name.
Generate a human-readable summary of a compiled build record.
Synchronize Java runtime to a specific version.
Validate a build against its contract specification.
Verify the validity of a terrain fit preview.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing what the response structure is. This violates the documented output schema requirement.
Six export_build_* tools (blueprint, schematic, litematic, mcfunction, csv, json) perform nearly identical operations with different output formats. This violates single-responsibility and forces LLMs to reason about which variant to use. Consolidate into a single export_build tool with a 'format' enum parameter.
No error handling guidance. Tools do not indicate which errors are retryable, which require user intervention, or how to recover. LLMs cannot self-correct or plan fallback strategies.
Destructive operations (delete_palette, install_world_edit_schematic, revise_selected_region) lack confirmation or dry-run support. No visible permission gates or audit logging. Agents could accidentally destroy data without safeguards.