Loading ToolBench
MCP server for Kratos Multiphysics: scaffold, run and post-process finite element simulations
Server has 25 tools with basic naming (verb_noun pattern mostly followed) and descriptions present. However, critical gaps emerge: most parameter descriptions are minimal (1-5 words), input schemas lack detail (no enums, ranges, or format constraints), and output schemas are entirely undocumented. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. Tool descriptions average ~40 chars, below the 194-char baseline for A+ tools. No tool annotations (readOnlyHint/destructiveHint) despite clear risk stratification (READ_ONLY vs WRITE). Composition is reasonable (tools are single-purpose), but chaining is weak, tools return job_id but downstream tools don't clearly document what fields they expect. Security concerns: kratos_install and run_simulation are high-risk WRITE operations with minimal parameter validation visible.
Add a boundary condition to a Kratos project
Configure checkpointing for a simulation
List available checkpoints for a job
Resume a simulation from a checkpoint
Scaffold a new Kratos finite element project
Export ProjectParameters to Flowgraph graph.json
Import Flowgraph graph.json to ProjectParameters
Cancel a running job
List all jobs
Output schemas completely undocumented. No tool documents what fields it returns, forcing LLMs to guess structure and breaking downstream tool chaining. E.g., job_status returns job_id but downstream tools don't document accepting it.
Parameter descriptions are minimal (1-5 words). E.g., 'case_dir' described as 'Project directory' lacks format, constraints, or examples. Rubric baseline is 72 chars; most params here are 10-20 chars.
No input validation constraints visible. Parameters like 'analysis_type' and 'bc_type' lack enums; 'frequency' and 'end_time' lack min/max bounds. LLMs will pass invalid values without guidance.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | <=2025-11-25 | v2 |
Get the status of a running or completed job
Introspect the Kratos installation
Install Kratos packages from PyPI
List available Kratos applications
List available Kratos variables
List available project templates
Convert mesh formats to/from MDPA
Create an MDPA mesh file
Inspect an MDPA mesh file
Create GIF animation of VTK results
Render PNG preview of VTK results
Summarize simulation results
Run a Kratos simulation as a background job
Run a mesh study
Run a parameter study
Validate a Kratos case configuration
No error handling or recovery guidance. Tools like kratos_install (WRITE, high-risk) and run_simulation provide no error messages, retryability hints, or next-step guidance if they fail.
No tool annotations despite clear risk stratification. 25 tools marked READ_ONLY or WRITE in metadata, but no readOnlyHint/destructiveHint/idempotentHint in schemas. Agents cannot infer safety properties.