claude-qa-mcp

Test documentation chases requirements across Miro, Jira, Confluence, Figma, and TestRail.
This config wires Claude Code into all of them — 2-3 days of docs land in 5-6 hours.

What Is This

claude-qa-mcp is a drop-in MCP server config for Claude Code, tailored for QA engineers. It connects the five tools where QA context lives — requirements, designs, tickets, test cases, automation — so Claude Code can read and write across all of them in a single session.

What it does:

Wires up 4 MCPs in one file — Miro, Atlassian (Jira + Confluence), TestRail, and Playwright
Triggers browser OAuth on first use for Miro and Atlassian — no API keys to manage
Uses 3 env placeholders for TestRail — URL, username, API key — that's it
Walks you through Figma separately via Claude AI's MCP catalog (the 5th MCP, not a drop-in)
Includes example prompts that exercise the full stack end-to-end

What You Get

| MCP | What it does for QA | Auth | |---|---|---| | Miro | Read flow diagrams, process maps | Browser OAuth on first use | | Atlassian (Jira + Confluence) | Pull tickets, specs, acceptance criteria | Browser OAuth on first use | | TestRail | Read / create / update test cases and sections | API key + URL + username (env vars) | | Playwright | Drive a real browser; draft test automation | None |

After install, /mcp inside Claude Code shows:

mcp servers
├── miro          ✓ ready    (browser OAuth on first call)
├── atlassian     ✓ ready    (browser OAuth on first call)
├── testrail      ✓ ready    (env credentials)
├── playwright    ✓ ready    (no auth)
└── figma         install separately — see below

Quick Start

1. Clone the repo

git clone https://github.com/demian-v/claude-qa-mcp
cd claude-qa-mcp

2. Drop the config — pick one

Global: merge the mcpServers block from .mcp.json into ~/.claude.json
Per-project: copy .mcp.json into your project root

3. Fill in the TestRail placeholders in your copy:

"TESTRAIL_URL": "https://yourcompany.testrail.io",
"TESTRAIL_USERNAME": "you@yourcompany.com",
"TESTRAIL_API_KEY": "your-api-key-here"

4. Restart Claude Code, then run /mcp to confirm all 4 servers are listed.

Figma (added separately)

Figma isn't in .mcp.json — there's no public HTTPS endpoint to drop in. Install it through Claude AI's MCP catalog:

Open https://claude.ai/mcp
Find Figma in the catalog, click Connect
Approve the OAuth scopes in your browser
Restart Claude Code; /mcp should now list figma

This adds Figma to Claude's hosted MCP layer — it works in Claude Code automatically alongside the servers in .mcp.json.

Common Workflows

Prompts that exercise the full stack:

"Read Jira ticket QA-123, pull the linked Figma frame, and draft TestRail cases
for the new checkout flow under section 'Payments / Checkout'."

"Look at this Miro board, find every payment branch, and create a TestRail
test case for each one in the 'Payment flows' section."

"Pull the acceptance criteria from this Confluence page, draft Playwright tests
covering the happy path, and save them under tests/checkout/."

The pattern: pull context from Miro/Atlassian/Figma → AI drafts → TestRail or Playwright lands the output.

How to get your TestRail API key

In TestRail, click your avatar → My Settings
Open the API Keys tab
Click Add Key, give it a name (e.g. claude-code)
Copy the generated key — it's only shown once
Paste it into .mcp.json as TESTRAIL_API_KEY

TESTRAIL_URL is your TestRail instance URL (e.g. https://yourcompany.testrail.io). TESTRAIL_USERNAME is the email you log in with.

Prerequisites

Claude Code
uvx for the TestRail MCP — install via pip install uv or brew install uv
node + npx for the Playwright MCP — comes with Node.js
A modern browser (for the Miro and Atlassian OAuth flows)

Atlassian Cloud vs Server

The official Atlassian Remote MCP works with Atlassian Cloud instances (*.atlassian.net). Self-hosted Atlassian Server / Data Center isn't supported by this remote endpoint — those need a different setup.

Honest Scope

This file gets you 4 of 5 MCPs working. Figma needs the catalog step above.
AI alone won't write the test cases you need. The MCP layer plus project-specific context (skills, memory plugins, memory tools) is where the time savings come from. More on that layer in the LinkedIn series.

Security

The .mcp.json in this repo contains placeholders only — no real credentials
Never commit your filled-in .mcp.json to a public repo
Add .mcp.json to your project's .gitignore if you drop it into a project root
TestRail API keys carry the same blast radius as your account — create a dedicated key, not a personal one

Project Structure

claude-qa-mcp/
├── .mcp.json     # 4 MCP servers — drop into ~/.claude.json or project root
├── README.md     # this file
├── .gitignore
└── LICENSE       # MIT

Tech Stack

Agent: Claude Code with MCP servers (/mcp)
Config format: JSON — mergeable into ~/.claude.json or per-project .mcp.json
Transports: HTTP (Miro, Atlassian) + stdio (TestRail via uvx, Playwright via npx)
Authentication: Browser OAuth (Miro, Atlassian, Figma) + API key in env (TestRail)

About the Author

I'm Demian — Senior QA Engineer with 5+ years in e-commerce SaaS, test automation, and AI-assisted workflows. I built this config from my own QA workflow and packaged the methodology so others can drop it in.

License

MIT — use freely, no warranty.

MCP Servers