From superlore
Guides users through deploying a superlore knowledge base via self-hosting on Vercel (free) or joining the superlore Cloud managed waitlist. Use when someone asks to deploy or host superlore docs.
How this skill is triggered — by the user, by Claude, or both
Slash command
/superlore:superlore-deployThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
superlore is open source and **self-hosts free today**. Managed hosting — **superlore Cloud** — is in
superlore is open source and self-hosts free today. Managed hosting — superlore Cloud — is in
private beta (waitlisted), so the one-command superlore deploy path is not generally available
yet. This skill does two things: (1) get the user on the Cloud waitlist, and (2) walk them
through self-hosting on Vercel (or anywhere) so they're live now. Deployment is always the
user's — superlore never hosts their data without their say-so.
Before deploying anything, re-check the auth posture — especially for a company KB:
superlore.json. If type: "company-kb" and auth.enabled is false, warn the user: a
public deploy makes the site and the MCP world-readable. Strongly recommend enabling Google
SSO before going live, and offer to flip it on (auth.enabled: true, provider: "google"). Only
proceed public if they reconfirm.superlore deploy (one-command managed hosting with a hosted MCP endpoint, SSO/governance, and team
workspaces) is in private beta. Don't pretend it's live. Instead:
apps/docs/app/cloud/page.tsx; on the
public site it's /cloud).What Cloud will add (for context, no prices — it's not selling yet): managed hosting + one-click updates with a hosted MCP endpoint, Google/SAML SSO + org gate + roles/audit, and shared team workspaces over one MCP URL.
This is the real, working path. superlore is a standard Next.js 16 + Fumadocs app, so it deploys to Vercel (or any Next host) like any Next app.
Build locally first to catch errors: superlore build (or npm run build / pnpm build).
Narrate the result.
Push to a Git repo the user controls (GitHub/GitLab/Bitbucket).
Import to Vercel — New Project → import the repo → Vercel detects Next.js. No special build config needed for a default superlore app.
Set environment variables if auth is enabled (auth.enabled: true in superlore.json). These
gate the site and the MCP via proxy.ts:
| Variable | Required | Purpose |
|---|---|---|
AUTH_GOOGLE_ID | to enable | Google OAuth client id — its presence turns auth ON |
AUTH_GOOGLE_SECRET | to enable | Google OAuth client secret |
AUTH_SECRET | yes (auth) | Secret used to encrypt the session (JWT) |
AUTH_URL | yes (auth) | The deploy's canonical URL |
AUTH_TRUST_HOST | yes (auth) | Set true behind a proxy / on Vercel |
AUTH_ALLOWED_DOMAIN | no | Restrict SSO to one workspace domain, e.g. acme.com |
AUTH_ALLOWED_EMAILS | no | Comma-separated email allowlist that bypasses the domain check |
With no AUTH_GOOGLE_ID, auth stays off and the deploy is fully public — that's the
intended public path.
Never ask the user to paste secrets into the chat. Tell them to add these in Vercel's project
settings (Environment Variables) or via vercel env add. If a value is missing, ask them to
set it themselves rather than guessing or inventing one.
Deploy — Vercel builds and gives a URL. The MCP comes up at <deploy>/api/mcp (or the
mcp.path from superlore.json).
Verify: open the site; then connect the MCP and confirm search / get_page work — hand off
to superlore-connect-mcp.
For non-Vercel hosts: it's a standard Next 16 app — next build then run/serve per the host's Next
guide. A static export works for a fully public KB with no server-side auth; the MCP route needs a
running server (or Cloud).
npx claudepluginhub krishnansg/superlore --plugin superloreGuides completion of development work by verifying tests, detecting environment, and presenting structured options for merge, PR, or cleanup.
Guides creation and editing of skills using test-driven development with pressure scenarios and subagents to verify agent compliance.
Dispatches multiple subagents concurrently for independent tasks without shared state. Use when facing 2+ unrelated failures or subsystems that can be investigated in parallel.