How to Use Cursor AI in VS Code to Auto-Generate Grafana SLO/SLI Dashboards from Prometheus Queries

Why Combine Cursor AI, VS Code, and Grafana in 2026?

Teams that run Prometheus at scale need clear SLO/SLI views. Building a Grafana dashboard by hand can take hours, especially when you have to copy PromQL, set thresholds, and align panels. In 2026, Cursor AI’s Composer mode can call external tools directly from the editor. When you add the official Grafana MCP server, Cursor can run PromQL, fetch metric metadata, and write dashboard JSON without leaving VS Code.

Real-world usage shows a 30-40% reduction in time-to-visibility for new services (Grafana 2026 user survey). The workflow also cuts manual copy-paste errors, because the AI writes the exact query you asked for.

Below you’ll see how to set up the stack, write effective prompts, and verify the generated dashboards.

Step 1 – Install Cursor AI and Enable MCP Support

Cursor AI ships as a desktop app for Windows, macOS, and Linux. Download the 2026 release from cursor.com and follow the installer prompts.

After installation, open Cursor Settings (Cmd+, on macOS, Ctrl+, on Windows/Linux). Scroll to **Features → MCP Servers** and click **Add New MCP Server**.

Enter the following JSON. It pulls the official Grafana MCP Docker image, which includes tools for Prometheus queries and dashboard CRUD operations:

{
  "grafana": {
    "command": "docker",
    "args": ["run", "--rm", "-i", "-e", "GRAFANA_URL=https://grafana.mycompany.com", "-e", "GRAFANA_TOKEN=${GRAFANA_TOKEN}", "grafana/mcp-grafana:2026.1"],
    "cwd": "."
  }
}

Save the file as .cursor/mcp.json in your project root. Toggle the switch to enable the server. Cursor will now show a **@** mention picker with tools like query_prometheus and create_dashboard.

Step 2 – Connect Your Prometheus Instance

The Grafana MCP server reads the GRAFANA_URL and GRAFANA_TOKEN you supplied. It also discovers any Prometheus datasource configured in Grafana. If you need a dedicated token for Prometheus, add it to the same environment block:

{
  "grafana": {
    "command": "docker",
    "args": ["run", "--rm", "-i", "-e", "GRAFANA_URL=https://grafana.mycompany.com", "-e", "GRAFANA_TOKEN=${GRAFANA_TOKEN}", "-e", "PROMETHEUS_URL=http://prometheus:9090", "grafana/mcp-grafana:2026.1"],
    "cwd": "."
  }
}

Test the connection by opening Cursor Composer, typing @query_prometheus, and entering a simple query such as up{job="api"}. The AI will show the raw result in the output pane.

Step 3 – Prompt Cursor to Generate an SLO/SLI Dashboard

Write a clear, concise prompt. Cursor works best when you give the AI a goal, the metric name, and the SLO target. Example:

"Create a Grafana dashboard named **API Availability SLO**. Use the Prometheus metric http_requests_total with label status="200". Define an SLO of 99.9% success over the last 30 days. Add a panel that shows the error rate as a single-stat, and a time-series panel for the success ratio. Export the dashboard JSON to a file called api-availability-slo.json."

When you send the prompt, Cursor’s Composer switches to Agent mode, calls query_prometheus to fetch the raw data, then calls create_dashboard with a generated JSON payload. The AI returns a trace like:

{
  "tool": "create_dashboard",
  "args": {"title": "API Availability SLO", "panels": [...]},
  "result": {"uid": "abc123", "url": "https://grafana.mycompany.com/d/abc123/api-availability-slo"}
}

After the trace, Cursor writes the JSON file you asked for. Open the file to verify the panels, thresholds, and variable definitions.

Step 4 – Review, Tweak, and Commit

Even though the AI builds the dashboard, you should still review the generated JSON. Look for:

• ✅ Correct PromQL expressions
• ✅ Proper thresholds field for the SLO target
• ✅ Panel titles that match your naming conventions

If you need a small change, edit the JSON directly in VS Code and run @update_dashboard with the dashboard UID. The AI will apply the diff without opening the Grafana UI.

When the dashboard passes review, add the JSON file to your Git repo. Grafana can import dashboards from source control using the grafana-cli dashboards import command, keeping monitoring as code.

Comparison Table – Cursor AI vs. Competing AI-Assisted Tools (2026)

So what does this mean? If you need write-back capability—creating or updating dashboards—Cursor AI offers the most seamless experience at a competitive price. Claude Code is a solid alternative for read-only analysis, while Copilot X still lacks direct MCP integration.

Practical Takeaways – Who Should Use This?

Site reliability engineers (SREs) who already own Grafana and Prometheus will find the workflow fastest. The AI removes the manual copy-paste step and lets you keep dashboards in code.

DevOps teams that manage many micro-services can generate a template dashboard for each service with a single prompt, then tweak variables per service.

Junior developers learning observability can see a live example of a correct PromQL query and how it maps to a Grafana panel, accelerating their learning curve.

Common Pitfalls and How to Avoid Them

1. Missing Grafana token scope. The MCP server needs Dashboard:Read and Dashboard:Write. Grant these scopes in the Grafana Cloud UI before you start.

2. Large PromQL results. Cursor’s context window can overflow if a query returns millions of points. Use range_query with a reasonable step (e.g., 1m) or add rate() to reduce data size.

3. Accidental writes in production. Enable the “Ask before write” toggle in Cursor Settings for production environments, or run the AI in a sandbox namespace first.

Conclusion – Turn Prometheus Data into Actionable SLO/SLI Dashboards in Minutes

Cursor AI in VS Code now lets you generate Grafana SLO/SLI dashboards from Prometheus queries with a single prompt. The 2026 Grafana MCP server provides reliable tool calls, and the AI’s Agent mode can write dashboards directly to your monitoring stack. By following the steps above, you cut dashboard creation time, reduce errors, and keep observability as code.

Start today: install Cursor, add the Grafana MCP server, and ask the AI to build your first SLO dashboard. Your monitoring team will thank you.