Use `gitinit` directly from AI coding agents.
The hosted MCP server exposes the same generator that powers the website. Agents can discover stacks, preview scaffold shape, generate a scaffold, and download the same zip artifact a human would get from the UI.
Artifact URLs are intentionally short-lived and should be fetched promptly by the calling agent.
Clients only need the hosted MCP endpoint: https://gitinit.arnayshukla.com/api/mcp.
If your agent supports instructions such as `AGENTS.md`, add a short rule telling it to use the `gitinit` MCP server whenever it needs to scaffold a new backend service.
Start with the generic remote MCP shape, then use the client-specific example that matches your coding agent.
{
"mcpServers": {
"gitinit": {
"url": "https://gitinit.arnayshukla.com/api/mcp"
}
}
}Cursor reads MCP config from `~/.cursor/mcp.json` or project-local `.cursor/mcp.json`.
{
"mcpServers": {
"gitinit": {
"url": "https://gitinit.arnayshukla.com/api/mcp"
}
}
}codex mcp add gitinit --url https://gitinit.arnayshukla.com/api/mcpCodex can also be configured directly through `~/.codex/config.toml`. The CLI and IDE extension share the same MCP config.
[mcp_servers.gitinit]
url = "https://gitinit.arnayshukla.com/api/mcp"Antigravity commonly manages MCP servers through a `mcp_config.json` file or its managed MCP editor.
{
"mcpServers": {
"gitinit": {
"serverUrl": "https://gitinit.arnayshukla.com/api/mcp"
}
}
}This is the same request shape the hosted MCP tool surface accepts for `preview_scaffold` and `generate_scaffold`.
{
"request": {
"stack": "java-spring-maven",
"projectName": "inventory-service",
"identifiers": {
"groupId": "com.example"
},
"selections": {
"database": "postgres",
"cache": "redis",
"messaging": "kafka",
"tracing": "opentelemetry",
"auth": "jwt",
"authorization": "rbac",
"rateLimit": "inMemory"
}
}
}