WIDEAREAINTELLIGENCE

for people who want to run their own ai

The models you choose, on hardware you own.

Pick any open-weight model. Run it on the machines you already have. Wide Area Intelligence pools them into one OpenAI-compatible endpoint you control — no vendor deciding what you can run, no per-token fees, and the cloud waiting only as failover.

Run your own AI — freeBrowse the model directory
[ any open-weight model ][ pool your machines ][ no vendor lock-in ]

/// own the stack

No vendor between you and your model

When you rent inference, a provider decides which models exist, when they get deprecated, and what every token costs. Run your own and those decisions come back to you: the weights you picked, on the hardware you bought, behind an endpoint you control.

Wide Area Intelligence handles the part that's normally annoying — making a box under your desk reachable, pooling several machines, and keeping a stable URL in front of them. A one-line install opens a secure tunnel; the gateway load-balances across every node you add and falls back to the cloud only when your hardware can't take the request.

/// who this is for

Built for people who'd rather own it

builders & teams

You want your own AI

You like the idea of running the models yourself — choosing the weights, owning the endpoint, answering to no vendor's roadmap. Pick any open-weight model and give it a real, reachable home on hardware you own.

research lab

CapEx beats a subscription

A recurring SaaS bill for a whole lab gets rejected; a one-time hardware grant gets approved. Pool the workstations and servers the lab already has into a shared, private AI sandbox your whole group can query.

distributed team

Pool the hardware you have

Your team already owns capable machines sitting idle most of the day. Combine their GPUs into one endpoint — capacity you've already paid for, working as a single shared pool instead of scattered boxes.

/// how the routing works

Your hardware first. The cloud only when you need it.

01metric 0 · ~10ms

EDGE CACHE

An identical request you've served before is returned straight from Cloudflare's edge — the prompt never reaches a GPU. Per-account toggle, your own TTL.

02metric 1 · no token fees

YOUR HARDWARE

A one-line install turns any machine with a GPU into a node. It opens a secure Cloudflare Tunnel — no port forwarding, no static IP — and serves llama.cpp behind your gateway with no per-token fees.

03metric 2 · always up

CLOUD FAILOVER

Node busy, offline, or timed out? The same request silently re-routes to the cloud through our managed gateway, billed only against prepaid credits — burst only, never the baseline.

/// drop-in integration

Your model, your endpoint.
Standard OpenAI calls.

Once your nodes are registered, your own models answer to a single OpenAI-compatible URL. Anything that already speaks the OpenAI API — SDK, LangChain, agents, coding tools, curl — runs on the hardware you own by changing one line: the base URL.

  • Change one line: the base URL.
  • Works with the OpenAI SDK, LangChain, agents, curl — anything OpenAI-compatible.
  • Bring your own gateway key; routing, caching, and failover are automatic.
app.py
from openai import OpenAI

client = OpenAI(
    base_url="https://wideareaai.com/api/v1",
    api_key="wai_sk_…",
)

resp = client.chat.completions.create(
    model="llama-3.1-8b-instruct",
    messages=[…],
)

/// go deeper

Set up your own AI

free toolVRAM calculatorCheck which models fit the GPUs you already own before you load them.free toolGPU for a modelFind the hardware to run the open-weight model you want to own.pillarExpose a local modelTurn an Ollama / LM Studio / llama.cpp box into a reachable endpoint.segmentKeep your data yoursOwning the hardware is also what keeps your prompts private.

/// faq

Questions about running your own

What does it mean to run my own AI?
It means you choose the open-weight model, it runs on hardware you own, and no vendor sits between you and your inference deciding what you can run, deprecating it, or metering every token. Wide Area Intelligence turns the machines you already have into nodes and pools them behind one OpenAI-compatible endpoint you control.
Which models can I run?
Any open-weight model your hardware can serve through llama.cpp — Llama, Qwen, GLM, Mistral, Gemma, DeepSeek, and the rest. You pick the model and the quantization; the gateway gives it a stable endpoint. Because you own the deployment, a model you depend on can't be retired out from under you.
Can I pool more than one machine?
Yes — that's the point of the 'wide area' in the name. A one-line install turns each machine with a GPU into a node, and the gateway load-balances across all of them behind a single endpoint. Add the desktop, the workstation, and the server you already own and they serve as one pool.
Do I lose access to frontier cloud models?
No. Running your own models is the default, and the cloud is still there as failover for the requests your hardware can't or shouldn't handle. You own the baseline and rent only the burst — control without giving up reach.
How hard is it to set up?
One line to install the node agent, which opens a secure Cloudflare Tunnel — no port forwarding, no static IP — and registers your hardware with the gateway. Your apps then talk to one OpenAI-compatible URL. Anything that speaks the OpenAI API works by changing the base URL.

/// initialize

Run the AI you choose. On the hardware you own.

Run your own AI — free

no credit card · 2 nodes free, forever · openai-compatible