MCP server for AI agents to interact with Grafana dashboards and datasources.
Visit WebsiteFreeVisit Website
TL;DR - Grafana MCP
- Connects AI agents to Grafana dashboards, datasources, and queries via the Model Context Protocol.
- Supports context-window-efficient operations like dashboard summaries, JSONPath property extraction, and patching.
- Works with any MCP client (Claude Desktop, Cursor, Copilot) and can be installed via uvx, Docker, or Helm.
Pricing: Free forever
Best for: Individuals & startups
What is Grafana MCP?
Grafana MCP is an open-source Model Context Protocol (MCP) server that provides AI assistants with direct access to Grafana instances. It enables AI agents to search, retrieve, update, and create dashboards, query datasources, and run panel queries — all through standardized MCP tooling. The server is designed to work with any MCP-compatible client, including Claude Desktop, Cursor, and GitHub Copilot.
Built for observability workflows, Grafana MCP implements context-window management strategies such as dashboard summaries, JSONPath property extraction, and patch operations to minimize token usage. It supports multiple installation methods (uvx, Docker, binary, Helm) and requires Grafana 9.0 or later for full functionality. The server also integrates with Grafana Cloud and supports datasource types like Prometheus, Loki, ClickHouse, and CloudWatch.
Pros & Cons
Pros
- Open source and free to use with no licensing restrictions
- Reduces context window consumption with dashboard summaries, property extraction, and patch operations
- Works with any MCP-compatible client, enabling integration with Claude, Cursor, Copilot, and more
Cons
- Requires Grafana 9.0 or later for full functionality; older versions may have limited datasource support
- Panel query tools are disabled by default and must be explicitly enabled via command-line flags
Preview
Key Features
Search for dashboards by title or metadataGet dashboard by UID, summary, or specific property via JSONPath expressionsCreate, update, or patch dashboards without full JSON payloadsRetrieve panel queries and datasource information (UID, type) for each panelRun panel queries with custom time ranges and variable overridesList and fetch detailed information about all configured datasourcesSupports datasource types including Prometheus, Loki, ClickHouse, and CloudWatchMultiple installation methods: uvx, Docker, binary, and Helm charts
Pricing Plans
Pricing checked Jul 31, 2026
Open Source
Free
- Full source code access
- Community support
- Self-hosted
Reviews
Be the first to review Grafana MCP
Your take helps the next buyer. Verified LinkedIn reviewers get a badge.
Write a reviewBest Grafana MCP Alternatives
Top alternatives based on features, pricing, and user needs.
DynatracePaid
Full-stack observability and AIOps platform
Elastic ObservabilityPaid
Full-stack observability solution built on a Search AI Platform, enabling faster troubleshooting with agentic AI.
DatadogFreemium
Cloud monitoring, security, and AI investigations for DevOps
New RelicFreemium
Full-stack observability with 50+ monitoring capabilities
SplunkPaid
Data platform for security and observability
PrometheusFree
Monitoring and alerting toolkit
Explore More
Grafana MCP FAQ
What is the minimum Grafana version required for Grafana MCP?
Grafana 9.0 or later is required for full functionality. Some features, particularly datasource-related operations, may not work correctly with earlier versions due to missing API endpoints.
How do I install Grafana MCP for use with Claude Desktop or Cursor?
Add the MCP server configuration to your client's settings. For Claude Desktop, include a JSON block with the command 'uvx mcp-grafana' and environment variables 'GRAFANA_URL' and 'GRAFANA_SERVICE_ACCOUNT_TOKEN'. Alternative installation methods include Docker, binary download, and Helm charts.
Which datasource types are supported by Grafana MCP?
The server supports Prometheus, Loki, ClickHouse, and CloudWatch datasources, and can retrieve datasource UID and type information from any configured datasource. Additional datasources may be accessible depending on your Grafana instance's plugins.
How can I reduce context window usage when working with large dashboards?
Use the 'get_dashboard_summary' tool for an overview, or 'get_dashboard_property' with a JSONPath expression to extract only the needed data. The 'patch dashboard' tool allows targeted modifications without sending the full dashboard JSON, significantly reducing token consumption.
How do I enable the 'run panel query' tool in Grafana MCP?
Run panel query tools are disabled by default. To enable them, add 'runpanelquery' to the '--enabled-tools' flag when starting the MCP server. This allows execution of a dashboard panel's query with custom time ranges and variable overrides.
Can I use Grafana MCP with Grafana Cloud instances?
Yes, replace the 'GRAFANA_URL' environment variable with your Grafana Cloud instance URL (e.g., https://myinstance.grafana.net) and provide a valid service account token. The server works with both self-hosted and cloud-hosted Grafana.
Does Grafana MCP support creating or modifying dashboards?
Yes, the server provides tools to update or create a dashboard (requires full dashboard JSON) and a 'patch dashboard' tool for applying specific changes without the full JSON. Both operations require appropriate API permissions on the Grafana instance.
How does Grafana MCP handle multiple concurrent sessions?
The server includes session management with horizontal scaling support, as indicated by files like 'session_horizontal_scaling_test.go' and 'session_reaper_test.go'. This allows handling multiple AI agent requests simultaneously with proper state isolation.
Source: github.com