Grafana Cloud MCP
Grafana Cloud MCP lets agents query metrics, logs, traces, dashboards, datasources, alerts, incidents, and related observability resources in an authorized Grafana Cloud stack.
Verdict
Common use cases
- Pull CPU metrics during incident postmortems
- Summarize alert history for weekly reviews
- Query logs to debug production errors
- Compare dashboard trends across deployments
- Fetch trace spans for latency investigations
Integration
- Vendor
- Grafana Cloud MCP
- Category
- other
- Auth
- OAUTH2
- Composio slug
grafana_cloud_mcp
Tools
Setup
Setup guide
- 11. Open your Switchy workspace and navigate to Settings > Integrations > Browse MCP Servers. 2. Search for Grafana Cloud MCP and click Connect. 3. You'll be redirected to Grafana Cloud's OAuth consent screen — sign in with your organization account. 4. Grant read access to dashboards, data sources, and alerting (the MCP requests viewer-level scopes only). 5. After authorization, Switchy confirms the connection and shows your connected Grafana Cloud stack name. 6. Open any Project and type '@Grafana Cloud MCP list my dashboards' to verify the connection works. 7. If you see a list of dashboards, you're ready — the MCP can now query metrics and pull observability data into any conversation.
What teammates see: by default, memories from Grafana Cloud MCP are scoped to the Project (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Projects have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Dashboard Summary
@Grafana Cloud MCP show me the last hour of metrics from the 'API Performance' dashboard and summarize any anomalies
Alert History Review
@Grafana Cloud MCP list all alerts that fired in the past 7 days, grouped by severity, and highlight any recurring patterns
Log Search
@Grafana Cloud MCP search logs for 'database timeout' errors in the checkout service over the last 24 hours
Metric Comparison
@Grafana Cloud MCP compare p95 latency for the payments API between yesterday and today, broken down by endpoint
Trace Lookup
@Grafana Cloud MCP pull the trace for transaction ID abc123 and show me which service added the most latency
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@Grafana Cloud MCP show me the current error rate and p95 latency for the checkout service over the last hour
Based on the metrics in your Grafana Cloud instance, the checkout service shows an error rate of 2.3% over the last hour, with a p95 latency of 340ms. The error rate spiked to 4.1% around 14:22 UTC, coinciding with a latency increase to 580ms. Current values (last 5 minutes) show error rate at 1.8% and p95 latency at 310ms, suggesting the issue is resolving.
This example demonstrates read access to time-series metrics. Requires OAuth2 scope for metrics query access. The MCP translates natural language into PromQL or LogQL queries behind the scenes. Response accuracy depends on your dashboard configuration and metric naming conventions in Grafana Cloud.
@Grafana Cloud MCP create an alert rule that notifies #incidents if API response time exceeds 1 second for more than 5 minutes
Alert rule created successfully in your Grafana Cloud instance. Rule name: 'API Response Time Threshold'. Condition: avg(http_request_duration_seconds) > 1 for 5m. Notification channel: Slack #incidents. The alert is now active and will evaluate every 60 seconds. You can view and modify this rule in the Grafana alerting UI under 'Alert rules'.
This write operation creates persistent infrastructure in your Grafana Cloud account. Requires OAuth2 scope for alert management. The MCP cannot delete or modify existing alerts without explicit confirmation. Be aware that alert creation counts toward your Grafana Cloud plan limits.
@Grafana Cloud MCP analyze the logs from the payment service in the last 30 minutes and identify any patterns in the errors we're seeing
Analyzing 847 log entries from payment-service over the last 30 minutes. Three error patterns identified: (1) 'timeout connecting to payment gateway' appears 23 times, clustered between 14:15-14:25 UTC. (2) 'invalid card token' occurs 8 times, evenly distributed. (3) 'rate limit exceeded' shows 5 occurrences, all from the same client IP. The timeout errors correlate with the latency spike visible in your metrics, suggesting an upstream gateway issue rather than application-level problems.
This synthesis example combines log querying with AI reasoning to surface insights. Requires OAuth2 scope for Loki log access. The AI interprets log patterns but cannot guarantee causal relationships—correlation analysis should inform, not replace, deeper investigation. Log retention and query performance depend on your Grafana Cloud tier.
Use-case deep-dives
When Grafana Cloud MCP fits on-call triage workflows
A 3-person SRE team gets paged at 2am for a latency spike. The on-call engineer needs to query metrics, check dashboards, and correlate logs without leaving the incident Slack thread. Grafana Cloud MCP is the right call here if your team already runs Grafana Cloud and the OAuth2 handshake is pre-configured in your workspace. The MCP can surface time-series data and dashboard snapshots directly in chat, cutting the context-switch tax. The trade-off: if your observability stack spans multiple vendors (Datadog, New Relic, CloudWatch), you'll still need separate integrations for each. This MCP wins when Grafana Cloud is your single source of truth and your team wants AI-assisted metric queries during live incidents.
Using Grafana Cloud MCP for recurring metric check-ins
A 5-person product team holds a Friday sync to review conversion funnels and API latency trends. The PM wants to pull last week's p95 latency and error rate without opening Grafana's UI. Grafana Cloud MCP works if the team's metrics live in Grafana Cloud and someone on the team has already set up OAuth2 for the workspace. The MCP can fetch time-series data and answer "what was our error rate on Tuesday" in natural language. The threshold: if your team only checks metrics once a week, the setup overhead might outweigh the benefit. This MCP is worth it when metric lookups happen multiple times per day and the team wants to keep those queries inside their shared AI workspace instead of bouncing to a browser tab.
When Grafana Cloud MCP helps support answer SLA questions
A 2-person support team fields customer questions about API uptime and response times. A customer asks "was there downtime on the 15th?" and the support rep needs to check without escalating to engineering. Grafana Cloud MCP is a fit if your SLA metrics live in Grafana Cloud and the support team has read-only OAuth2 access. The MCP can answer point-in-time queries and surface uptime percentages in seconds. The boundary: if your support team doesn't have direct Grafana Cloud access today, adding OAuth2 for them might be overkill unless they field metric questions daily. This MCP wins when support handles 5+ metric lookups per week and you want to give them self-service access without full Grafana training.
Frequently asked
What does the Grafana Cloud MCP do in Switchy?
The Grafana Cloud MCP connects your observability data to Switchy's AI workspace. Your team can query metrics, logs, and traces through natural language instead of writing PromQL or LogQL. It pulls data from your Grafana Cloud stack so AI agents can answer questions about system performance, error rates, and infrastructure health without leaving the conversation.
Does Grafana Cloud MCP require OAuth2 setup?
Yes, this integration uses OAuth2 authentication. You'll authorize Switchy to access your Grafana Cloud organization through a standard OAuth flow. The specific scopes depend on which Grafana Cloud features you want to expose — typically read access to metrics, logs, and dashboards. You don't need to manage API tokens manually; the OAuth connection handles credential refresh automatically.
Can the Grafana Cloud MCP create or modify dashboards?
That depends on the OAuth scopes you grant during setup. Most teams start with read-only access to query existing data and dashboards. If you need AI agents to create alerts or modify dashboard configurations, you'd grant write scopes — but read-only is safer for most use cases. Check the scope selection screen during OAuth authorization to see what's available.
How does this compare to querying Grafana Cloud directly?
Querying Grafana Cloud directly requires knowing PromQL, LogQL, or navigating the UI to build queries. The MCP lets non-technical team members ask questions in plain English and get answers from your observability data. You trade some query precision for speed and accessibility. For complex investigations, engineers will still use Grafana's native interface.
Who on my team should connect the Grafana Cloud MCP?
Someone with admin or editor permissions in your Grafana Cloud organization should handle the initial OAuth connection. After that, any Switchy team member can query the connected data through AI agents. If your Grafana Cloud plan has user limits, the MCP connection typically counts as one service account or OAuth app, not per-user seats.