Loading ToolBench
Local stdio MCP server for CONVERGE Studio automation
Server has 11 tools with explicit schemas and descriptions. Naming is mostly verb-driven (studio_*, solver_*) and clear. Descriptions are present and substantive (avg ~120 chars), exceeding the 10-char minimum. Input schemas use Zod with type definitions and constraints. However, several tools lack parameter descriptions (solver_list, solver_read, solver_search lack descriptions for individual params like 'caseDirectory', 'path', 'pattern'). Output schemas are not documented, responses are JSON but structure is not formally declared. Error handling exists (StudioMcpError, errorPayload) but recovery guidance is minimal. Tool composition is sound (each does one thing), but some tools combine multiple concerns (e.g., studio_execute handles both working-copy and source mutations with conditional gates, increasing cognitive load).
Recursively list files inside an allowlisted raw solver case. Symbolic links are reported but never followed.
Read a bounded line range or tail from a text input/log file inside an allowlisted raw solver case.
Regex-search bounded text files inside an allowlisted raw solver case. Files over the configured limit and symbolic links are skipped.
Inspect raw solver-case allowed roots, persistent job directory, safety gates, and limits without reading a case or launching a process.
Parse an allowlisted ASCII surface.dat and report overall plus per-boundary triangle counts, unique vertex counts, coordinate bounds, spans, and vertex centroids.
List all available CONVERGE Studio bridge commands with their parameters and descriptions.
Parameter descriptions missing or incomplete. solver_list, solver_read, solver_search have Zod schemas with type info but lack human-readable descriptions for 'caseDirectory', 'path', 'pattern', 'maxDepth', 'maxEntries', etc. LLMs cannot infer parameter meaning from names alone.
Output schemas not documented. Tools return JSON via resultPayload() but the structure of returned fields is not formally declared. LLMs cannot plan downstream calls or extract data reliably without knowing response shape.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
Query MCP tool coverage for CONVERGE Studio bridge commands and CSI methods.
Execute a CONVERGE Studio bridge command with optional working-copy or source mutation modes.
Execute a GUI-driven workflow in CONVERGE Studio with optional parameters.
Retrieve detailed manual documentation for a CONVERGE Studio bridge command or CSI method.
Inspect MCP configuration, discovered Studio/bridge/CVG paths, catalog version, and execution queue without launching Studio.
studio_execute combines multiple concerns: working-copy vs source mutation modes, optional gates, optional working-copy path, optional CVG file. Conditional logic increases complexity and error surface. Consider splitting into studio_execute_working_copy and studio_execute_source.
Error recovery guidance is minimal. errorPayload() exists but error messages in tool implementations do not guide LLMs on next steps (e.g., 'Try search_users() first' or 'Retryable: yes/no').
No pagination support. solver_list, solver_search, studio_catalog return results without limit/offset or next_cursor. Large result sets could exhaust context window.