MCP server for injection molding simulation knowledge — material properties, process validation, troubleshooting, and simulation spec generation
MoldSim MCP demonstrates solid tool design with clear naming conventions, good descriptions, and well-structured input schemas using Zod validation. All 6 tools follow verb_noun naming patterns and have substantive descriptions (194-350 chars average). Parameter schemas are properly typed with Zod constraints. However, the server lacks documented output schemas, error recovery guidance, and tool annotations (readOnlyHint). All tools are read-only but this is not explicitly declared via annotations. The knowledge base query tool relies on embedding models without exposing failure modes or recovery paths. Descriptions could be more concise (several exceed 300 chars) and output structure is not documented, forcing LLMs to infer response format.
Compare 2-4 materials side-by-side. Returns a table of processing, thermal, and mechanical properties with key differences highlighted. Available materials: abs-generic, pp-homo, pp-copo, pa6, pa66, pa66-gf30, pc, pc-abs, pom, hdpe, ldpe, pmma, pbt, pbt-gf30, pet, ps, hips, tpu, san, asa
Generate a Design for Manufacturability (DFM) checklist for an injection molded part. Returns pass/warn/fail status for 15+ design rules based on part description and parameters.
Generate a structured simulation specification from a natural language description of the part and requirements. Outputs analysis types, process conditions, mesh recommendations, and expected results.
Look up material properties for injection molding simulation. Returns Cross-WLF viscosity model, 2-domain Tait PVT, thermal properties, processing window, and mechanical data. Available materials: abs-generic, pp-homo, pp-copo, pa6, pa66, pa66-gf30, pc, pc-abs, pom, hdpe, ldpe, pmma, pbt, pbt-gf30, pet, ps, hips, tpu, san, asa
Search injection molding simulation knowledge base. Covers troubleshooting (defects, simulation issues), DFM rules, mesh guidelines, and process parameter recommendations. Use for general questions about molding, simulation setup, or problem-solving.
No documented output schemas. All tools return 'text' content type with unstructured responses. LLMs cannot plan downstream tool chains or extract structured data. Violates pattern:response-shaper.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 6 tools are read-only but LLMs cannot determine this from the schema, forcing conservative retry logic.
query_simulation_knowledge tool relies on embedding model (HuggingFace transformers + Qdrant) without documented error handling. No recovery guidance if embedding model fails, warm-up hangs, or vector DB is unavailable.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 62 | 2026-07-28+ | v2 |
| 2026-03-09 | C | 60 | - | v1 |
Validate injection molding process parameters against material processing window. Checks melt/mold temperature, estimates shear rate, calculates cooling time, and flags out-of-range values with suggestions.
Material lookups accept ambiguous input ('pa6', 'PA6', 'pa66-gf30', 'PA66-GF30'). No documented validation, normalization, or case-sensitivity rules. Tool descriptions list all 20 valid materials inline, wasting tokens and risking hallucination of invalid material names.
Descriptions exceed 300 characters for generate_simulation_spec (348) and generate_dfm_checklist (331), exceeding the 200-char baseline for optimal LLM processing. Excess detail dilutes key intent signals.