jowls 🍒
165 posts
jowls 🍒
@itsjowleebee
cherry dust magic to all!
Introducing @Laureum_ai — quality scoring for MCP servers and AI agents by @assisterr We score 6 dimensions: accuracy, safety, reliability, process quality, latency, and schema quality. Multi-judge LLM consensus + adversarial probes. We've scored 28 public MCP servers to date. Average: 68.3/100. 6 in Expert tier (≥85). The weakness nobody else measures: process quality — averaging 55.5/100. Here's why we built it👇 Three gaps in agent eval today: → Marketplaces curate by hand. A major MCP catalog operator pruned 17 abandoned /vanity / impersonation entries from their own catalog earlier this month — manually. → Eval frameworks (LangSmith, Braintrust, Galileo) score tool-call correctness well. Process quality — error handling, input validation, response structure — sits between them, and nobody surfaces it as a named composite. → Post-Drift, the Solana ecosystem just launched STRIDE for smart-contract security. Agent infra still ships without pre-deploy quality gates. Laureum is the missing layer. Free right now, no signup: 1/ Quick Scan — paste any MCP server URL, get a 30-second 6-axis score → laureum.ai/evaluate 2/ Public leaderboard — see how the most-used servers rank → laureum.ai/leaderboardIf you're building, run yours. Reply with your score — we'll feature the top 5 this week. End of the tweet.
