A11ychk MCP server
{
"mcpServers": {
"a11ychk": {
"command": "npx",
"args": [
"-y",
"@a11ychk/mcp@latest"
]
}
}
}{
"mcpServers": {
"a11ychk": {
"command": "npx",
"args": [
"-y",
"@a11ychk/mcp@latest"
]
}
}
}{
"mcpServers": {
"a11ychk": {
"command": "npx",
"args": [
"-y",
"@a11ychk/mcp@latest"
]
}
}
}{
"servers": {
"a11ychk": {
"command": "npx",
"args": [
"-y",
"@a11ychk/mcp@latest"
]
}
}
}{
"mcpServers": {
"a11ychk": {
"command": "npx",
"args": [
"-y",
"@a11ychk/mcp@latest"
]
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y @a11ychk/mcpversion 0.1.4
What it does
A11ychk is a web accessibility scanner that checks pages against WCAG 2.2 and KWCAG 2.2 (Korea's national guidelines), then generates Korean-language remediation guides and fix requests for developers. It serves Korean-speaking accessibility practitioners and development teams who need structured audit reports and AI-assisted code fixes.
- Run accessibility checks on representative pages from a website URL and get dual-mapped findings against both WCAG 2.2 and KWCAG 2.2 standards with Korean improvement guides.
- Generate automated remediation documents in Markdown or JSON that can be fed directly into AI coding tools like Cursor or Copilot to assist with fixing violations.
- Integrate accessibility gating into CI/CD pipelines via GitHub Actions to fail builds when serious-level violations are detected before deployment.
Tools this server exposes
| Tool | Does |
|---|---|
| audit | Scans a URL for accessibility violations using axe-core plus custom rules, maps findings to WCAG 2.2 and KWCAG 2.2, and returns structured results with Korean r |
| kwcag_audit | Performs a specialized accessibility audit focused on KWCAG 2.2's 33 checkpoints with Korea-specific remediation guidance. |
| get_rule_guide | Retrieves the Korean improvement guide and AI coding instructions for a specific accessibility rule. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.