Grafana Cloud¶
What it is¶
Grafana Cloud is a fully managed observability platform that provides unified monitoring for metrics, logs, traces, and application performance. It includes hosted versions of Prometheus, Loki, Tempo, and Grafana, along with specialized AI Observability features for LLM-powered applications.
What problem it solves¶
It centralizes monitoring from disparate sources into a single dashboarding interface. For AI applications, it enables tracking of LLM latency, token usage, and error rates alongside traditional infrastructure metrics. Its Actually Useful AI™ suite, including Grafana Assistant, helps automate incident analysis and dashboard generation.
Where it fits in the stack¶
Infrastructure / Observability / Eval. It serves as the primary visualization and alerting layer for the OpenTelemetry ecosystem.
Typical use cases¶
- Multi-source Dashboards: Combining AWS CloudWatch, Prometheus, and LLM logs into one view.
- AI Agent Monitoring: Tracking 95th percentile operation duration and cost attribution for agentic systems.
- Log Aggregation: Using Loki to search through distributed agent logs with trace correlation.
- VectorDB Observability: Monitoring query performance and resource utilization for vector databases.
Strengths¶
- Open Standard Support: Native support for Prometheus and OpenTelemetry.
- Rich Visualization: Industry-leading dashboarding capabilities.
- AI-Powered Insights: Built-in agents for root cause analysis and incident summaries.
- Scalability: Managed infrastructure handles high volumes of telemetry data.
Limitations¶
- Complexity: Setting up advanced dashboards and alerts requires significant knowledge of PromQL or LogQL.
- Data Silos: Requires active effort to ensure all relevant data is being ingested.
- Public Preview: Some AI Observability features are still in public preview as of June 2026.
When to use it¶
- When you already use Grafana for infrastructure and want to add AI observability.
- When you need high-performance, long-term storage for logs and metrics.
- When you want to leverage MCP to manage dashboards and query data via AI assistants.
When not to use it¶
- For simple applications where basic logging is sufficient.
- If you prefer a purely local, lightweight observability stack.
Getting started¶
Installation¶
Grafana Cloud doesn't require a local installation for the UI, but you typically need an agent like Grafana Alloy to ship data.
# Install Grafana Alloy (example for Debian/Ubuntu)
sudo apt-get install alloy
Basic Setup¶
- Create a free account at grafana.com.
- Navigate to AI Observability in the sidebar to enable the public preview features.
- Configure your LLM providers (e.g., Claude or OpenAI) to ship OpenTelemetry data to your Grafana endpoint.
CLI examples¶
Using Grafana Alloy to ship logs¶
# Start Grafana Alloy with a local config
alloy run config.alloy
Querying Loki logs via LogCLI¶
# Query logs for a specific job
logcli query '{job="varlogs"}' --addr="https://logs-prod-us-central1.grafana.net"
Managing Dashboards via Grafana CLI¶
# List installed plugins
grafana-cli plugins ls
API examples¶
Shipping LLM Metrics with OpenTelemetry (Python)¶
Grafana Cloud supports OpenTelemetry natively. You can use the OpenTelemetry SDK to track token usage for models like Claude 4.8 or GPT-5.5.
from opentelemetry import metrics
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
API_URL = "https://otlp-gateway-prod-us-central1.grafana.net/v1/metrics"
API_TOKEN = "your_grafana_cloud_token"
headers = {
"Authorization": f"Basic {API_TOKEN}"
}
exporter = OTLPMetricExporter(endpoint=API_URL, headers=headers)
# ... configure meter and instrument to track token usage
Querying Loki via API¶
import requests
API_URL = "https://<your_loki_user>:<your_loki_api_key>@logs-prod-us-central1.grafana.net/loki/api/v1/query"
params = {'query': '{job="agent-logs"}'}
response = requests.get(API_URL, params=params)
print(response.json())
Related tools / concepts¶
- Datadog - Enterprise-grade observability and security.
- New Relic AI - Full-stack AI monitoring and alerting.
- OpenTelemetry Collector - Vendor-neutral telemetry proxy.
- Model Context Protocol (MCP) - Standard for connecting agents to data.
- Claude - Frontier LLM for orchestration.
- LlamaIndex - Data framework for LLM applications.
- Prometheus - Standard for k8s monitoring.
- Loki - Horizontally scalable log aggregation.
- Tempo - High-volume distributed tracing.
Sources / references¶
- Grafana AI Observability Documentation
- Grafana MCP Server
- Actually Useful AI™ in Grafana Cloud
- Llama 4 Maverick Observability Patterns
Contribution Metadata¶
- Last reviewed: 2026-06-28
- Confidence: high