RAGScore
Verified · yesterdayGenerate QA datasets & evaluate RAG systems with failure diagnosis. Any LLM.
Install
claude mcp add ragscore -- uvx ragscore
Our take
RAG evaluation in two commands: generate QA pairs from your documents, fire them at your RAG endpoint, and get accuracy scores with per-question failure diagnosis. Works with any LLM including local Ollama, which keeps evaluation private, and the notebook API with built-in plots makes it genuinely quick to audit a pipeline. Young project from a new org — the methodology is LLM-as-judge with all the caveats that implies, so read the scored failures rather than trusting the single headline number.
Verification record
- yesterday
- 22
- 38 days ago
- no
- Apache-2.0
- 2026-05-19