Hey Product Hunt! I built Dominion Observatory because I was frustrated that MCP directories only show stars and install counts — nothing about whether a server actually works reliably. We now track 14,824 MCP servers with behavioral trust scores based on runtime data: latency, error rates, success rates, response consistency. Key findings: - Average trust score: 64.5/100 (Grade C) - Top servers: under 50ms latency, 98%+ success rate - Bottom quartile: high errors, stale maintenance You can check any server for free — just curl our API or browse the directory. We also have PyPI packages for FastMCP, LangChain, CrewAI, and OpenAI Agents SDK that add trust checks before every MCP tool call. Would love your feedback — what trust signals matter most when you're picking MCP servers for your agents?

Dominion Observatory

Behavioral trust scores for 14,824 MCP servers

Behavioral trust scores for 14,824 MCP servers