Svgrid MCP server
{
"mcpServers": {
"svgrid": {
"command": "npx",
"args": [
"-y",
"@svgrid/mcp@latest"
]
}
}
}{
"mcpServers": {
"svgrid": {
"command": "npx",
"args": [
"-y",
"@svgrid/mcp@latest"
]
}
}
}{
"mcpServers": {
"svgrid": {
"command": "npx",
"args": [
"-y",
"@svgrid/mcp@latest"
]
}
}
}{
"servers": {
"svgrid": {
"command": "npx",
"args": [
"-y",
"@svgrid/mcp@latest"
]
}
}
}{
"mcpServers": {
"svgrid": {
"command": "npx",
"args": [
"-y",
"@svgrid/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 @svgrid/mcpversion 3.0.1
What it does
An MCP server that validates AI-generated SvGrid code against the real API, demo source code, and version-pinned Svelte 5 documentation. It is for developers using Claude or other AI agents to write SvGrid data grid code, ensuring generated code matches the actual API before it reaches the developer.
- Verify that AI-generated SvGrid column definitions, props, and component usage match the actual API surface.
- Retrieve correct example code from 370+ demos when the AI agent needs reference implementations.
- Check that generated code uses Svelte 5 syntax and SvGrid patterns, returning exact corrections for wrong props or column keys.
Tools this server exposes
| Tool | Does |
|---|---|
| validate_svgrid_code | Checks AI-written SvGrid code against the real API surface and returns corrections for invalid props, column keys, or syntax. |
| get_demo_source | Retrieves source code from the 370+ SvGrid demos for reference and example patterns. |
| check_svelte5_syntax | Verifies that generated code uses Svelte 5 runes and syntax rather than Svelte 4 patterns. |
| query_api_docs | Searches the version-pinned Svelte 5 SvGrid documentation for property definitions, types, and usage guidance. |
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.