Loading ToolBench
RunBox MCP server — code execution tool for Claude Desktop, Cursor, and MCP-compatible agents
RunBox MCP has 3 tools with complete input schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names are action-oriented (execute_code, discover) but descriptions lack LLM-optimization guidance. The execute_code_with_files tool has a complex nested schema but minimal guidance on file format expectations. No error handling guidance, no output schema documentation, and missing parameter constraints (e.g., code length limits, timeout behavior). Descriptions average ~150 chars but lack WHEN/WHY context. No tool annotations (readOnlyHint/destructiveHint) despite execute_code being a WRITE operation.
Discover RunBox capabilities, supported languages, pricing, and payment info.
Execute code in an isolated Docker container via RunBox. Supports: python, javascript, bash, ruby, php, perl, lua, go, rust, java, c, cpp, typescript, r. Code runs in a sandboxed environment with no network access. Requires a valid session token (from Stellar USDC payment).
Execute code with file input/output. Files are mounted at /input/ (read) and /output/ (write). Send files as base64, get output files back as base64.
Missing output schema documentation. No tool returns a documented schema, LLMs cannot plan downstream calls or extract fields reliably.
No error handling guidance. Descriptions do not explain what errors can occur, whether they are retryable, or what the LLM should do next (e.g., 'Invalid session token, call discover() to check payment status').
Missing parameter constraints and descriptions. 'code' parameter has no length limit, timeout, or format guidance. 'language' enum is present but no description of what happens if unsupported language is passed.
No tool annotations. execute_code and execute_code_with_files are destructive (WRITE risk) but lack destructiveHint annotation. LLMs cannot distinguish safe from unsafe tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
Incomplete parameter documentation. execute_code_with_files.files array has nested 'content' field described as 'Base64-encoded' but no guidance on max file size, supported encodings, or error handling for decode failures.