An overview of Rippin Host — what the portal does, how billing and provisioning work,
and what we shipped as of , including
game server hosting and bare-metal dedicated servers housed with a colocation carrier.
Language barrier is no more
Translate the whole site from the top nav — English, Spanish, and many more short-code languages.
What’s new
The portal and marketing site can now be read in your preferred language. We use
Google Translate behind a simple language control in the top navigation,
so the UI is no longer English-only for customers in Europe, India, Latin America, and beyond.
How to change language
Look at the top-right of the navigation bar (after your account menu when
signed in, or after Sign in / Get started when signed out).
Open the language control — it shows a globe and a short code such as
EN, ES, or DE.
Pick your language. The page translates in place (or reloads once if needed). Your choice
is remembered for later visits in this browser.
To go back to English, open the same menu and choose EN.
Available short codes include EN, ES, PT, DE, FR, NL, PL, IT, HI, ID, VI, TR, JA, KO,
ZH, AR — Spanish, Portuguese, German, French, Hindi, Indonesian, and others side by
side with English.
Notes
Translation is automatic (machine). Menus, buttons, and help text are covered; product names
such as OpenClaw, n8n, and Hermes stay as-is. Dashboard values (container names, URLs,
ticket subjects you typed) are left as you entered them. If something looks off, switch back
to EN and tell us via a support ticket.
Network path: where we are, and where we’re going
Honest status on uplink, on-prem edge routers, interim bandwidth, and the plan to scale.
Today’s path
Our site has a 1 Gbit physical last-mile connection, but public traffic
is not riding that full capacity yet. Connectivity is currently
terminated over a WireGuard VPN to a far-end MikroTik
RB1100 Dude Edition, with an RB2011 in the edge mix.
Under that path, practical uplink sits around
~50 Mbit/s — enough for a young platform, not what we will settle for.
Near-term routing upgrade
We’re working to bring online another MikroTik
RB1100AHx2 or RB1100AHx4 so we can stop bottling the path
through the current VPN/edge arrangement. With that in place, we expect to use on the order
of ~150–200 Mbit/s of the existing 1 Gbit circuit — a large step
for containers, Shared Ollama API traffic, and customer dashboards without waiting on a full
carrier rebuild.
Offload, not heroics
In parallel, we intend to move more customer-facing services onto rented capacity
rather than packing everything through this single on-prem uplink. That keeps latency and
headroom under control while the edge gear and carrier quote catch up. You shouldn’t feel
like you’re living on our lab bandwidth forever.
Longer-term: dedicated Frontier capacity
We have not yet received formal pricing from our uplink provider
(Frontier) for a dedicated or shared-dedicated jump toward
10 Gbit class service. That may read small next to megascale hosts —
we know — but it is the honest roadmap: grow the circuit, upgrade the routers, and keep
workloads on the right side of the fence. We’ll post again when hardware lands or a quote
turns into a real install window.
OpenClaw × Shared Ollama, manage services, and hardening
Proxy compatibility for OpenClaw, multi-service management, and safer signups.
Shared Ollama works better with OpenClaw
OpenClaw talks to Shared Ollama over the OpenAI-compatible path
(/v1/chat/completions using your personal base URL
(https://ollama.rippinhost.com/rh_ollama_… (set that as
OLLAMA_BASE_URL — do not append /v1).
Agents often send function tools in every chat request. Base
llama3:latest on our host does not support tool calling, so Ollama was
returning HTTP 400 and sessions failed even for ordinary chat. The authenticated proxy now
strips unsupported tool fields for those models so completions succeed. Prefer
llama3:latest unless you need another pulled model. Native tool-calling still
requires a tools-capable model on the GPU host; chat and agent text turns work through the
proxy today.
Create or copy your base URL any time from
Ollama access.
Manage services
Dashboard “Manage services” now covers add and remove for
your subscription — pick plans carefully on each tab. OpenClaw cards also remind you when a
provider key or OLLAMA_BASE_URL still needs to be set under Variables.
Safer registration
New accounts include address fields, hCaptcha, per-IP limits, attempt rate limits, honeypot /
timing checks, and disposable-email blocking — less signup abuse without blocking legitimate
customers.
Shared Ollama on NVIDIA M40
Shared GPU inference runs on an NVIDIA M40 today; Tesla T4 upgrade is planned later.
Current hardware: M40
Shared Ollama is live on our NVIDIA M40 GPU host. API keys, base URLs, and the
$9.95/mo flat rate stay the same —
no per-token billing. n8n and OpenClaw keep using your personal
Ollama proxy URL; prefer llama3:latest unless you need
another pulled model.
We still plan a move to dual NVIDIA Tesla T4 cards later, but that cutover is
not imminent. We’ll post again when the T4 host is ready and the switch is
scheduled. Your keys and base URLs will carry over when it happens.
Web Hosting (invite-only)
Classic shared hosting — $10/mo, 20 domains, unlimited disk space on DirectAdmin.
Beyond containers
We are introducing Web Hosting alongside our Docker catalog: shared hosting
with a DirectAdmin control panel, email, DNS, and SSL — up to 20 domains and
unlimited disk space for $10/mo.
Public self-serve signup stays off for now. The plan appears on the
homepage as invite-only; customers receive a private
registration link to create an account and subscribe.
Container plans (n8n, OpenClaw, Hermes, and the rest) are unchanged — Web Hosting is a separate
flat add-on for traditional sites and email.