A Model Context Protocol server for Python development tools
This MCP server provides 12 tools covering file operations, code analysis, and project management. Tool definitions are present with schemas and descriptions, but significant quality gaps prevent a higher score. Schemas are properly structured with parameter types and descriptions. However, many descriptions are generic and lack the specificity needed to guide LLM selection. Error handling is minimal, there is no evidence of recovery guidance, categorization of retryable vs fatal errors, or validation feedback. Several high-risk tools (execute_python, delete_file, write_file) lack proper safeguards or confirmation patterns. Tool composition has issues: write_file and delete_file are destructive operations that should support dry-run or confirmation. The execute_python tool is particularly dangerous without sandboxing details or execution limits. Naming is generally clear (verb_noun pattern), but descriptions average ~60 chars, below the 194-char production baseline and too brief to fully guide LLM behavior. Output schemas are not documented, so LLMs cannot plan downstream calls or extract specific fields.
Analyze the structure of a Python file
Check for dependency conflicts
Create a virtual environment
Delete a file
Execute Python code
Format Python code according to style guidelines
Install project dependencies
execute_python tool lacks sandboxing, timeout, and execution limits documentation. No mention of resource constraints, allowed modules, or failure modes. This is a critical security and stability gap for an arbitrary code execution tool.
Destructive tools (delete_file, write_file) lack confirmation or dry-run patterns. No guidance in descriptions about consequences. Agents may accidentally delete critical files without safeguards.
No documented output schemas. For tools like analyze_python_file, list_packages, and check_conflicts, LLMs cannot determine what fields to expect in responses. This prevents effective downstream tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Run linting on Python code
List contents of a directory
List installed packages
Read the contents of a file
Write content to a file
Descriptions are generic and lack WHEN/WHY context. E.g., 'Analyze the structure of a Python file' doesn't explain use cases (refactoring? documentation? testing?). Average description length ~60 chars vs production baseline ~194 chars. Descriptions should state what problem the tool solves and when to prefer it over alternatives.
No error handling or recovery guidance. Tools provide no documentation on what errors may occur, which are retryable, or what the LLM should do next. E.g., execute_python could fail with SyntaxError, ImportError, timeout, or resource exhaustion, none are documented.
Parameters lack constraint documentation. E.g., 'mode' in write_file accepts 'w' or 'a' but this is not documented in the description. 'style' in format_code accepts 'black' or 'pep8' but no description explains which to use when.
list_directory path parameter has a default of '.' but no description of what relative paths mean. Does it resolve relative to workspace_root or current working directory? Ambiguous defaults invite misuse.
create_venv and install_dependencies have no input parameters visible but accept optional ones (e.g., requirements_file). If a tool genuinely has no params, the schema should be empty {} with clear documentation. If params exist, they must be documented.