rm-rf-prod/GroundTruth-MCP
Self-hosted MCP server for live docs and best practices. 330+ libraries, no rate limits, no API keys, code audit tool.
Platform-specific configuration:
{
"mcpServers": {
"GroundTruth-MCP": {
"command": "npx",
"args": [
"-y",
"GroundTruth-MCP"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
<p align="center"> </p>
<h1 align="center">GroundTruth</h1>
<p align="center"> Your AI assistant just confidently wrote deprecated code again.<br/> This fixes that. </p>
<p align="center"> <a href="https://www.npmjs.com/package/@groundtruth-mcp/gt-mcp"></a> <a href="https://github.com/rm-rf-prod/GroundTruth-MCP/actions/workflows/ci.yml"></a> <a href="./LICENSE"></a> </p>
---
They hallucinate APIs. Not because the models are bad — because documentation moves faster than training data does. A library ships a major version, deprecates half its API, and the model keeps generating the old patterns for the next six months. You catch it in review if you're lucky. In production if you're not.
There's also the security problem nobody talks about enough. AI assistants produce insecure patterns without knowing they're doing it: SQL built with template literals, innerHTML fed user input, any scattered through TypeScript like confetti, cookies() called without await in Next.js 16. The model isn't malicious — it just learned from code that predates the rule change, and now it's your problem.
Context7 helps with the first problem. But it has rate limits, live
Loading reviews...