A hosted Model Context Protocol server that gives Claude Code, Claude Desktop, Copilot or Cursor the package index, setup guides, worked examples and exact type signatures of the Regira packages. Generated code targets the real API instead of a guess. No install, no key, no sign-up.
https://mcp.regira.com/mcpThe server runs on its own host; a plain GET on mcp.regira.com shows an info page.
Streamable HTTP transport — nothing runs on your machine. Each client has its own way in; the JSON form is the same for the ones that take a file.
One command, or commit the JSON below as .mcp.json at the repo root. Approve the server the first time it asks; claude mcp list shows it connected.
Customize → Connectors → + Add → Add custom connector, name regira, paste the endpoint. No config file — on Team and Enterprise an Owner adds it once under Organization settings → Connectors.
VS Code reads the same .mcp.json at the workspace root, so one committed file serves both. Trust the server when it starts, then enable the Regira tools under Configure Tools in the Chat view.
The same JSON in .cursor/mcp.json in your project, or ~/.cursor/mcp.json for all of them.
# Claude Code — one command, then approve the server when asked
claude mcp add --transport http regira https://mcp.regira.com/mcp{
"mcpServers": {
"regira": {
"type": "http",
"url": "https://mcp.regira.com/mcp"
}
}
}The "type": "http" line is not optional: Claude Code and VS Code skip a url entry without it.
The server carries a bootstrap guide with a phased path — classify entities, host, register, prove the API, then the Vue front-end — each phase with a checkpoint. Name the domain and point the agent at it.
Scaffold a Regira Entities API for a shopping list with categories and articles.
Start from the Regira MCP server's bootstrap guide and follow its checkpoints.| Proof | Nine sample apps — API plus SPA — were generated end to end this way, each with a README on what the agent decided and where it deviated. |
|---|---|
| What it cost | 27–48 MCP calls and half an hour to two hours per app. Those are the agents' own tallies from their session transcripts — approximations, not instrumented measurements. |
| Where the calls go | Mostly documentation reads rather than code — that is the point: the agent checks signatures instead of guessing them. The SPA is generated slice by slice from @regira/modules and then customised by hand where the generic overview doesn't fit. |
| Free tier | Regira Entities is free up to 5 simple + 2 complex entity registrations per application; the agent counts yours before it designs and says so when the domain won't fit. The server itself allows 30 requests per minute per client without a key. One license can cover both. |
A dozen tools: package discovery, full-text search over every guide, per-package index cards, section-by-section reading, code examples, type signatures from the source map and task recipes. Connect it and ask your agent what it sees.
What the agent builds on: Entities, Office, IO & Storage and the rest.