Infrastructure & deployment
What a Random Draw from the MCP Registry Contains, and What Tool-Use Benchmarks Contain Instead
A seeded random draw found that server startup—not credentials—was the main obstacle, while tool duplication varied sharply across benchmark corpora.
Why it matters: Sampling and repair choices can make an MCP ecosystem look more operational than a random draw does. Benchmark statistics can also count repeated rows instead of distinct tools unless releases are globally deduplicated.