glossary / tool reliability
Tool reliability is whether an AI agent tool (an MCP tool or API) is working correctly right now — measured from real-usage success rate, latency, and error patterns, not just whether it is reachable.
A neutral, cross-ecosystem reliability rating lets an agent decide whether to call a tool, tune timeouts/retries, or switch to a live alternative — avoiding wasted calls on broken tools.
MCP server · MCP gateway · pre-flight check · working recipe · reliability score · schema drift
Glimind is the live reliability layer for AI agent tools — docs · add it.