DevGuard AI: A Self-Observing Multi-Agent Security Pipeline Built on SigNoz

Short summary

DevGuard AI is an autonomous multi-agent security pipeline that detects vulnerabilities, patches them, and adversarially reviews its own fixes using a Scanner-Fixer-Validator loop. It instruments every step with OpenTelemetry spans in SigNoz and closes the loop by reading its own telemetry via SigNoz's MCP server to adjust routing and enforce cost budgets in real time. The system includes custom dashboards for token cost, latency percentiles, cache efficiency, and circuit-breaker state, plus alert rules for cost overruns and SLO degradation.

• •Three-agent pipeline (Scanner, Fixer, Validator) with up to 3 retry loops for security patch validation
• •Full OpenTelemetry instrumentation in SigNoz with custom metrics for cost, latency, cache, and circuit-breaker state
• •Pipeline reads its own telemetry via SigNoz MCP server to dynamically route to cheaper models and enforce cost budgets

Generated with AI, which can make mistakes.