grub-crawler MCP: add nuts_services_list tool advertising N.U.T.S. endpoints

[kordless]

Summary

Add a nuts_services_list tool to MCP/grub-crawler.py that returns the known N.U.T.S. service directory — URLs, capabilities, auth requirements, and health status for each service.

Motivation

Agents using the grub-crawler MCP have no way to discover what other N.U.T.S. services exist or how to reach them. A discovery tool closes that gap and lets agents self-route to the right service (e.g. shivvr for ephemeral compute, auth for token exchange).

Proposed tool

@mcp.tool()
async def nuts_services_list(
    check_health: bool = False,
    ctx: Context = None,
) -> Dict[str, Any]:
    """
    List all known N.U.T.S. (Network Utilities and Tool Services) endpoints.

    Returns the service directory for the nuts.services fleet: URLs, purpose,
    auth requirements, and optionally live health status for each service.

    Args:
        check_health: If True, ping each service /health endpoint and include status.
        ctx: MCP context (optional).
    """

Services to advertise

Service	URL	Description
GrubCrawler	https://grub.nuts.services	Web crawler — clean extraction, bot bypass
Shivvr	https://shivvr.nuts.services	Ephemeral compute environments
Auth	https://auth.nuts.services	N.U.T.S. OAuth token exchange

Notes

• Static directory is fine initially; check_health=True pings /health on each
• Auth token (WRAITH_AUTH_TOKEN) should be passed through to services that require it
• GRUB_BASE_URL override should still apply for self-hosted deployments