Practical notes: Using k6 and Grafana MCP for AI-Powered Performance Testing

Use this as an operator-facing rebuild of the ideas in “Using k6 and Grafana MCP for AI-Powered Performance Testing”: clear stages, ordered code slots, and recovery notes that survive a handoff. The Overview stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

What the k6 MCP Actually Provides

For the What the k6 MCP stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary.

Setting Up k6 MCP

For the Setting Up k6 MCP stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary.

k6 x mcp

k6 x agent init cursor

k6 x agent init claude-code
k6 x agent init vscode-copilot
k6 x agent init codex-cli

mcp-k6

docker pull grafana/mcp-k6:latest

{
  "mcpServers": {
    "k6": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "grafana/mcp-k6"
      ]
    }
  }
}

First Use Case: Generate a k6 Test From a Requirement

For the First Use Case Generate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary. For the First Use Case Generate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

import http from 'k6/http';
import { check } from 'k6';

export const options = {
  stages: [
    { duration: '2m', target: 100 },
    { duration: '5m', target: 100 },
    { duration: '1m', target: 0 },
  ],  thresholds: {
    http_req_duration: ['p(95)<500'],
    http_req_failed: ['rate<0.01'],
  },
};export default function () {
  const response = http.get(
    'https://staging.example.com/api/products/search?q=laptop'
  );  check(response, {
    'status is 200': (r) => r.status === 200,
  });
}

Second Use Case: Validate Before Running a Real Load Test

When working through the Second Use Case Validate stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

Generate
   ↓
Validate
   ↓
Review
   ↓
Small execution
   ↓
Controlled load test

Generate
   ↓
"Run it"
   ↓
Surprise

Now Add Grafana MCP

When working through the Now Add Grafana MCP stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

Setting Up Grafana MCP

When working through the Setting Up Grafana MCP stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours. When working through the Setting Up Grafana MCP stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

{
  "mcpServers": {
    "grafana": {
      "command": "uvx",
      "args": [
        "mcp-grafana"
      ],
      "env": {
        "GRAFANA_URL": "https://your-grafana.example.com",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "YOUR_TOKEN"
      }
    }
  }
}

brew install mcp-grafana

Give Grafana the Right Permissions

The Give Grafana the Right stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

Fourth Use Case: Correlate k6 Results With Prometheus

The Fourth Use Case Correlate stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

http_req_duration p95: 1.2s
http_req_failed: 3.4%

k6
p95 latency ↑
      │
      ├── CPU ↑
      ├── DB connections ↑
      └── HTTP 5xx ↑

Fifth Use Case: Find the Error Behind a Performance Regression

The Fifth Use Case Find stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve. The Fifth Use Case Find stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Open Grafana
 ↓
Find Loki
 ↓
Find service
 ↓
Find time range
 ↓
Write query
 ↓
Inspect logs
 ↓
Compare with k6 timestamps

A Practical MCP Configuration

For the A Practical MCP Configuration stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary.

{
  "mcpServers": {
    "k6": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "grafana/mcp-k6"
      ]
    },
    "grafana": {
      "command": "uvx",
      "args": [
        "mcp-grafana"
      ],
      "env": {
        "GRAFANA_URL": "https://your-grafana.example.com",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "${GRAFANA_SERVICE_ACCOUNT_TOKEN}"
      }
    }
  }
}

AI coding agent
                       │
             ┌─────────┴─────────┐
             │                   │
          k6 MCP             Grafana MCP
             │                   │
             ▼                   ▼
        k6 execution       Prometheus / Loki
             │                   │
             └─────────┬─────────┘
                       ▼
                 AI investigation

the Recommended QA Workflow

For the the Recommended QA Workflow stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary.

Phase 1 — Generate

For the Phase 1 Generate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary. For the Phase 1 Generate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Requirement
     ↓
AI
     ↓
k6 script

Phase 2 — Validate

When working through the Phase 2 Validate stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

k6 script
     ↓
Validation
     ↓
Fix obvious errors

Phase 3 — Run safely

When working through the Phase 3 Run safely stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

1 VU
 ↓
small duration
 ↓
review

Phase 4 — Observe

When working through the Phase 4 Observe stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours. When working through the Phase 4 Observe stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

k6
 ↓
Prometheus
Loki
Dashboards

Phase 5 — Investigate

The Phase 5 Investigate stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

Performance failure
       +
Infrastructure telemetry
       +
Application logs
       ↓
Investigation

Phase 6 — Automate

The Phase 6 Automate stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

Where next ….

The Where next stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve. The Where next stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

BONUS — Useful Prompts

For the BONUS Useful Prompts stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Generate a test

For the Generate a test stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary.

Validate

For the Validate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Authenticate at the gateway and re-authorize at the data plane. A bearer token alone is not a tenancy boundary. For the Validate stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Investigate a failed test

When working through the Investigate a failed test stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

Correlate with metrics

When working through the Correlate with metrics stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours.

Investigate logs

When working through the Investigate logs stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Log tool name, args hash, latency, and outcome for every call. Debugging agent loops without that trail wastes hours. When working through the Investigate logs stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Compare runs

The Compare runs stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

Produce a report

The Produce a report stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Expose tools with narrow schemas and explicit side-effect labels. Hosts need to know which calls mutate state before they auto-approve.

Operational checklist

When working through the Operational checklist stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest.

Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.