Remote OpenClaw Blog
Best Observability MCP Servers in 2026
7 min read ·
Netdata is the best observability MCP server in 2026: the netdata/netdata project holds 79,403 GitHub stars as of July 2026, the most in the Remote OpenClaw directory's 341-server Observability category, and its built-in MCP server lets an agent query real-time metrics, logs, and anomalies across your infrastructure without a separate data pipeline. Which server is right for you depends on the stack you already run, so the seven picks below cover metrics dashboards, error tracking, time-series queries, and LLM observability, each ranked by GitHub stars with a verified install method.
How We Ranked the Best Observability MCP Servers
Observability MCP servers are Model Context Protocol servers that let an AI agent query monitoring data, metrics, logs, traces, and error reports, through natural language so it can investigate incidents and answer questions about system health. This list ranks them by GitHub stars in the Remote OpenClaw Observability category, verified against each repository on July 4, 2026, and picks winners across the main observability jobs rather than stacking near-identical metric tools.
One honesty note on stars. Netdata's 79,403 stars belong to the whole monitoring agent, not only its MCP interface, so we flag project-wide counts inline. For a workflow-focused companion, see our monitoring and observability MCP roundup.
The 7 Best Observability MCP Servers in 2026
These seven servers cover full-stack monitoring, dashboards, error tracking, time-series queries, LLM observability, and open-source APM. Install commands were verified against each repository's README in July 2026.
1. Netdata - real-time metrics and logs
Netdata (79,403 stars, counted for the whole project) exposes a built-in MCP server on every agent and parent node from v2.6.0, so an agent can query per-second metrics, logs, and machine-learning anomaly detection across your fleet. Connect through the nd-mcp bridge with the API key Netdata generates on the node. Directory page: netdata/netdata.
claude mcp add netdata /usr/sbin/nd-mcp --bearer YOUR_NETDATA_API_KEY ws://YOUR_NETDATA_IP:19999/mcp2. Grafana MCP - dashboards, Prometheus, and Loki
The official Grafana MCP (3,180 stars) queries dashboards, datasources, Prometheus and Loki queries, and incidents from chat, which makes it the best single server if Grafana is already your window into everything. Point it at your instance with GRAFANA_URL and a service account token. Directory page: grafana/mcp-grafana.
claude mcp add grafana -- docker run -i --rm -e GRAFANA_URL -e GRAFANA_SERVICE_ACCOUNT_TOKEN docker.io/grafana/mcp-grafana:0.10.03. Sentry MCP - error tracking and stack traces
The official Sentry MCP (741 stars) pulls error issues, stack traces, and release health into an agent, and its Seer integration can suggest root causes. It is the fastest way to let an agent triage production exceptions. Directory page: getsentry/sentry-mcp.
claude mcp add sentry -- npx -y @sentry/mcp-server4. Prometheus MCP - direct time-series queries
Prometheus MCP Server (483 stars) runs PromQL queries and inspects targets and metadata straight against a Prometheus instance, which suits teams that use Prometheus without Grafana on top. Set PROMETHEUS_URL and run it in Docker. Directory page: pab1it0/prometheus-mcp-server.
docker run -i --rm -e PROMETHEUS_URL="http://your-prometheus:9090" ghcr.io/pab1it0/prometheus-mcp-server:latest5. Opik MCP - observability for LLM apps
Opik MCP (211 stars) plugs an agent into Comet's Opik workspace to read LLM traces and spans, log evaluation scores, and version prompts, which is the observability layer for AI applications rather than servers. Directory page: comet-ml/opik-mcp.
claude mcp add opik -- uvx opik-mcp@latest6. SigNoz MCP - open-source APM and traces
SigNoz MCP (97 stars) queries the OpenTelemetry-native SigNoz platform for traces, metrics, and logs, giving an agent access to a full open-source APM stack. Provide SIGNOZ_URL and SIGNOZ_API_KEY and run it in Docker. Directory page: SigNoz/signoz-mcp-server.
docker run -p 8000:8000 -e SIGNOZ_URL="https://your-signoz.example.com" -e SIGNOZ_API_KEY="your-key" signoz/signoz-mcp-server:latest7. Prometheus alternatives and OpenTelemetry
For OpenTelemetry-first stacks, traceloop/opentelemetry-mcp-server (192 stars) exposes OpenTelemetry data to an agent through a vendor-neutral interface, which is useful when you route traces to multiple backends. It rounds out the list for teams standardizing on OTel rather than a single vendor. Browse more in the Observability directory.
claude mcp add otel -- npx -y @traceloop/opentelemetry-mcp-serverComparison Table: Best Observability MCP Servers at a Glance
The table below compares all seven picks on job, stars, and credentials, with star counts current as of July 4, 2026.
| Rank | Server | Best for | Stars (Jul 2026) | Credentials | Runs |
|---|---|---|---|---|---|
| 1 | Netdata | Real-time metrics and logs | 79,403 (project) | Node API key | Local bridge |
| 2 | Grafana MCP | Dashboards and datasources | 3,180 | Service account token | Local or Docker |
| 3 | Sentry MCP | Error tracking | 741 | OAuth or auth token | Remote or local |
| 4 | Prometheus MCP | PromQL queries | 483 | Prometheus URL | Docker |
| 5 | Opik MCP | LLM observability | 211 | Opik API key | Local |
| 6 | SigNoz MCP | Open-source APM | 97 | SigNoz URL + key | Docker |
| 7 | OpenTelemetry MCP | OTel-first stacks | 192 | OTel endpoint | Local |
What About Datadog?
Datadog has no widely adopted MCP server in the Remote OpenClaw directory as of July 2026. The community options, such as TANTIOPE/datadog-mcp-server, sit in the low single-digit star range and are early-stage, so we did not rank one. If you run Datadog and want agent access today, the most reliable path is a thin custom server against the Datadog API, or routing your metrics through Grafana and using the Grafana MCP above.
This is a fast-moving space, and an official Datadog server could land at any time, so check the live Observability directory for the current leader rather than trusting a static list. The directory re-ranks by stars automatically as new servers gain traction.
Limitations and Tradeoffs
Observability data is sensitive, and read tokens still leak information. Metrics, logs, and traces can contain hostnames, user identifiers, and error payloads, so scope every service account or API key to read-only and to the narrowest project the agent needs. Sentry and Grafana both support scoped tokens; use them. Our MCP security guide covers token scoping in detail.
Stars measure popularity, not fit. Netdata's 79,403 stars belong to the whole monitoring agent, and the best observability server for you is almost always the one that speaks to the platform you already run, not the highest-starred project overall. That is why the table pairs stars with the specific job each server does.
Related Guides
- Best MCP Servers for Monitoring and Observability
- Best MCP Servers for Claude Code in 2026
- Securing Your MCP Server Connections
- Best MCP Servers in 2026: The Complete Ranked List
Go deeper
The operator playbooks
Production-ready PDF guides for OpenClaw and Hermes Agent — $19.99 each.
Skills for this topic
Browse all skills →Frequently Asked Questions
What is the best observability MCP server in 2026?
Netdata is the best observability MCP server in 2026, with 79,403 GitHub stars as of July 2026 across the project, because it ships a built-in MCP server that exposes real-time metrics, logs, and anomaly detection with no extra pipeline. For dashboards the best pick is Grafana MCP, and for error tracking it is the official Sentry MCP.
Is there a Datadog MCP server?
There is no widely adopted Datadog MCP server as of July 2026; the community options are early-stage and low-star, so we did not rank one. If you use Datadog, a thin custom server against the Datadog API or routing metrics through Grafana MCP are the more reliable paths today.
Can an AI agent investigate incidents with an MCP server?
Yes. With an observability MCP server connected, an agent can query metrics, pull error stack traces, and correlate logs to help investigate an incident, then summarize likely causes. Netdata, Grafana, and Sentry all expose the read tools an agent needs for this, though a human should confirm any action before it is taken.
How do I install an observability MCP server in Claude Code?
Run claude mcp add <name> -- <command> , for example claude mcp add sentry -- npx -y @sentry/mcp-server , then provide the required token as an environment variable. Docker-based servers like Grafana and Prometheus MCP run through docker run . Our Claude Code MCP guide covers scopes and secrets.