Two days ago I measured that 4. 4% of MCP servers changed their tool contract in 36 hours , and refused to annualise it on the grounds that changes probably cluster. I now have a third snapshot, and the caution was warranted more strongly than I expected.
Source: [Dev.to](https://dev.to/theopslog/mcp-schema-drift-isnt-a-rate-its-a-small-set-of-servers-that-never-stop-moving-243c)