Kerq builds trust layer for AI tools

Kerq targets a real gap in agent infrastructure, but reliability scores alone cannot prove that a tool is safe or authorized.

• –Scores tools across uptime, latency, errors, consistency, throughput, recovery, and performance trends
• –Gives agent developers a machine-readable preflight signal for tool selection and fallback routing
• –Extends trust evaluation beyond models to the rapidly expanding MCP and agent-tool ecosystem
• –A healthy endpoint can still be malicious, over-permissioned, or vulnerable to prompt injection
• –Developers should combine Kerq with allowlists, scoped credentials, sandboxing, and runtime policy enforcement