ai-validation-mcp
v0.1.2
Published
MCP Server for independent AI code validation - analyzes tests for bias, missing edge cases, and coverage gaps
Maintainers
Readme
AI Validation MCP Server
Independent AI code validation service — analyzes your tests for bias, missing edge cases, and coverage gaps. Works as an MCP server for AI coding agents and IDEs.
Why
AI-generated code increases production incidents 43% YoY. General-purpose assistants generate tests as a side effect, not as a core function. This server provides specialized, independent validation — a second opinion on your test quality.
Installation
npm install ai-validation-mcpMCP Configuration
Add to your MCP client config (e.g. claude_desktop_config.json):
{
"mcpServers": {
"ai-validation": {
"command": "npx",
"args": ["-y", "ai-validation-mcp"],
"env": {
"ANTHROPIC_API_KEY": "your-key-here"
}
}
}
}Requires an Anthropic API key for analysis.
Tools
validate_tests
Analyze existing tests for bias, missing edge cases, and quality issues.
| Parameter | Type | Required | Description | |------------|--------|----------|--------------------------------------| | sourceCode | string | Yes | The source code to validate against | | testCode | string | Yes | The existing test code to analyze | | language | string | No | Programming language (default: typescript) |
Returns: bias warnings, missing edge case suggestions, overall quality score.
suggest_edge_cases
Suggest comprehensive edge cases for a given function.
| Parameter | Type | Required | Description | |------------|--------|----------|--------------------------------------| | sourceCode | string | Yes | The source code to analyze | | language | string | No | Programming language (default: typescript) |
Returns: categorized edge cases (happy path, error handling, boundary, type coercion, integration).
coverage_gap_analysis
Identify untested code paths, branches, and logic gaps.
| Parameter | Type | Required | Description | |------------|--------|----------|--------------------------------------| | sourceCode | string | Yes | The source code to analyze | | testCode | string | Yes | The existing test code to evaluate | | language | string | No | Programming language (default: typescript) |
Returns: coverage score, gap details, suggested tests to close gaps.
Modes
--cached(default) — Uses saved responses from./fixtures/. No API calls.--live— Real Anthropic API calls. Overwrites fixtures.
Development
npm install
npm run build
npm testLicense
MIT
