Model Context Protocol
MCP Servers Guide
Model Context Protocol servers for QA automation and AI-assisted testing workflows.
Playwright MCP Server
Playwright MCP server that gives AI agents full browser automation power through structured accessibility snapshots — no screenshots, no vision models. Pure structured data, maximum reliability.
GitHub MCP Server
GitHub MCP server (github/github-mcp-server) with 18 toolsets: issues, pull_requests, repos, actions, code_security, dependabot, projects, and more. Remote (OAuth) or local Docker stdio. Read-only and lockdown modes for safety.
Atlassian (Jira) MCP Server
Atlassian Remote MCP Server (Rovo, GA Feb 2026) — Jira, Confluence, and Compass with 72+ tools. OAuth 2.1 with Dynamic Client Registration (no app to register). Permissions respect existing Jira project roles.
MCP Servers for QA — Side-by-side
Pick the right server for the job. All three are actively maintained and free to run.
| Feature | 🎭 Playwright | 🐙 GitHub | 🪐 Jira |
|---|---|---|---|
| Transport | Local stdio | Remote HTTP (OAuth) or Docker stdio | Remote HTTP (OAuth 2.1) |
| Auth | None (local) | OAuth or PAT | OAuth 2.1 (DCR) or API token |
| Read-only mode | Inherent (browser only) | ✓ --read-only flag | ✓ scope-based via OAuth |
| Tools count | ~25 (navigate, click, type, snapshot) | 150+ across 18 toolsets | 72+ across Jira/Confluence/Compass |
| Best QA use case | E2E exploration, test codegen | PR triage, release notes, code review | Ticket triage, AC generation, sprint reports |
| Risk if write enabled | 🟢 Sandbox (own browser) | 🔴 Can close issues, merge PRs | 🟡 Can transition issues, post comments |
What is MCP?
The Model Context Protocol is an open standard (released by Anthropic in late 2024, GA across most clients in 2026) that lets AI assistants talk to external tools through a uniform interface. Instead of writing a custom integration for every IDE × every tool, you build one MCP server and every MCP-capable client (Claude Code, Cursor, Claude Desktop, Windsurf, VS Code Copilot, Zed, JetBrains, Gemini CLI) can use it.
For QA engineers, this means your AI assistant can drive a real browser (Playwright MCP), read PR diffs and post review comments (GitHub MCP), and create Jira tickets with acceptance criteria (Jira MCP) — from the same chat, without copy-pasting context.