SigNoz MCP: Agent Observability for SRE Teams
SRE teams are connecting AI agents to production telemetry. That can save time during incidents, but it also creates a governance problem. Agents need enough context to help, without broad, unmanaged access to every observability backend. SigNoz is interesting because it combines OpenTelemetry-native observability with an MCP server built for agent access.
What Is SigNoz?
SigNoz is an open-source observability platform built on OpenTelemetry. It brings logs, metrics, traces, alerts, dashboards, exceptions, infrastructure monitoring, and APM into one system. The project also supports LLM observability for prompts, tool calls, tokens, latency, and cost.
The newer operational angle is agent-native telemetry access. The SigNoz MCP server exposes observability data to tools such as Claude, Cursor, Copilot, Codex, Gemini CLI, and other MCP clients. For SREs, that means an agent can ask targeted questions about an incident instead of guessing from pasted log snippets.
Key Features
- OpenTelemetry-native storage for logs, metrics, traces, and application telemetry.
- MCP access for agents so assistants can query telemetry through a documented interface.
- LLM observability for prompts, RAG pipelines, tool calls, latency, usage, and cost.
- Kubernetes and infrastructure views for clusters, pods, nodes, workloads, hosts, and services.
- Self-hosted and cloud options for teams with different data residency and control requirements.
Installation
SigNoz supports cloud and self-hosted deployments. For self-hosting, the docs list Docker, Docker Swarm, Linux, Kubernetes, Argo CD, OpenShift, ECS, and other paths.
For Docker, the current docs use Foundry to generate and run the Compose stack:
curl -fsSL https://signoz.io/foundry.sh | bash
foundryctl cast -f casting.yaml
Foundry can also enable the MCP server on port 8000:
curl -fsS localhost:8000/livez && echo " OK"
Cloud users can configure a hosted MCP endpoint:
codex mcp add signoz --url https://mcp.<region>.signoz.cloud/mcp
codex mcp login signoz
Replace <region> with the region for your SigNoz Cloud account.
Incident Workflow
A useful first workflow is read-only incident investigation. Connect the agent to SigNoz, then ask narrow alert-driven questions:
Show traces for checkout errors in the last 30 minutes.
Compare p95 latency before and after the latest deployment.
Find logs correlated with spans from service payment-api.
This is better than copying dashboard screenshots into a chat window. The agent can inspect live context and help responders move from symptom to hypothesis faster.
Operational Tips
Start with service-account credentials scoped for investigation, not remediation. Keep MCP configuration out of version control if it contains API keys. Log every agent query, especially during incidents.
For Kubernetes teams, pair SigNoz with existing OpenTelemetry Collector pipelines first. Instrument a few critical services, validate trace-to-log correlation, then add MCP access once the telemetry model is stable.
Conclusion
SigNoz is worth watching because it treats AI-assisted operations as part of the observability system. MCP gives agents a standard way to read telemetry, while OpenTelemetry keeps signals portable.
Looking to automate infrastructure operations? Akmatori helps SRE teams reduce toil with AI agents built for real production workflows. For reliable global infrastructure, check out Gcore.