AI Gateway
AI Gateway: every AI call your business makes, logged and capped
Most businesses have no idea what their AI tools cost until the invoice lands, or what was sent to them at all. A gateway in front of every provider fixes both.
With Cascadia you get
Nothing installed on your machines
Costs you can see
Caps that hold
Set up and watched by us
Your providers, your keys
You own the logs
Why teams start here
Nothing installed on your machines
The only change is the endpoint your applications call. There is no agent to install and no code to rewrite.
Costs you can see
Caching stops you paying twice for the same answer, spend caps stop a runaway script costing hundreds overnight, and the log tells you which application is responsible.
Caps that hold
Prompts and responses pass through Cloudflare’s infrastructure, which we will walk you through before anything is switched on. If that is not acceptable, we will tell you rather than sell it to you anyway.
What you get
Everything AI Gateway covers
All of these come with AI Gateway, set up and looked after by our team.
Logging
- Every request to every provider, logged
- The prompt, the reply and the model recorded
- Which application made the call
- Tokens and cost per request
- Kept where you can search it
Cost control
- Caching, so the same answer is not paid for twice
- Spend caps you set, enforced at the gateway
- Rate limits per application
- Told before a budget is reached, not after
- Cheaper models used where they do
Reliability
- Fallback when a provider goes down
- One endpoint instead of five
- Provider outages ride through
- Retries handled for you
- No code change when a provider changes
Reporting
- A monthly report of what was spent
- Broken down by application
- What changed since last month
- Spend measured against the caps you set
- 90 days of history
What sets it apart
- Nothing installed on your machines
- Costs you can see
- Caps that hold
- Set up and watched by us
- Your providers, your keys
- You own the logs
Want to see what your AI actually costs?
Tell us what you sell and we will start with the questions your buyers ask.
Get startedSee the comparisonsCalling providers directly, and putting a gateway in front
Nothing stops you calling each provider directly, and plenty of businesses do. The difference shows up afterward: in what you can see, and in what happens when a script runs away or a provider goes down.
| Calling providers directly | AI Gateway | |
|---|---|---|
| Cost | Known when the invoice arrives | Logged per request, per application |
| Caps | None | Spend caps and rate limits you set |
| Logs | Whatever the provider shows | Prompt, reply, model and cost, kept 90 days |
| Outages | Your application stops | Fallback to another provider, retries handled |
| Caching | Paid for twice | Cached, so the same answer is paid for once |
| Upkeep | Set once | Watched and reported on every month |
What clients say about working with Cascadia
Want to see what your AI actually costs?
Talk to us
Who AI Gateway fits best
Teams whose AI bill is a surprise
You find out what the month cost when the invoice arrives, and never find out what was sent. Every request is logged as it happens, so the number stops being a surprise.
Applications calling more than one provider
One endpoint sits in front of all of them, with fallback and retries handled for you, so an outage at one provider does not take your application down with it.
Agencies running AI work for clients
You can sell AI cost control without building it. We run the gateway under your brand, hand you a monthly report broken down by application, and stay out of the relationship.
How it works
1. We put the gateway in front
We start by putting the gateway in front of the providers your applications already use. No code changes beyond an endpoint, and nothing installed on anyone’s machine.
2. We set the caps and watch it
Your applications are pointed at the gateway instead of the provider, which is an endpoint change and nothing more. Caching, rate limits and spend caps are set with you before anything is switched on.
3. You get the picture every month
Each month you get a plain report: what was spent, by which application, what changed, and anything unusual worth knowing about.
Ask us
Answers to common AI Gateway questions
Weighing us against another option? Our comparisons take the main ones in turn.
What is an AI gateway?
A gateway sits between your applications and the providers they call, so every request can be logged, cached, capped and retried without changing the applications themselves.
How is this different from reading each provider’s dashboard?
A provider dashboard shows you its own traffic and nothing else. The gateway sits in front of all of them, so the prompt, the reply, the model and the cost land in one place, with the application that made the call named.
Is this just Cloudflare’s AI Gateway?
Yes. The gateway is Cloudflare’s, and its core features are free. What you are paying for is it being set up properly, watched, and reported on rather than left as a dashboard nobody opens.
Do my applications need code changes?
Your applications point at our gateway endpoint instead of the provider directly. Your provider keys stay yours, and you can point them back at any time.
How long before I see results?
Setup starts within two business days, and the first full picture arrives within a week.
Can I see what was actually sent, and what it cost?
Yes. Logs are kept for 90 days in your portal, and you can see the prompt, the reply, the model, the cost and which application made the call.
What is in the monthly report?
What was spent, broken down by application, what changed since last month, and anything unusual worth knowing about. Ninety days of history sits behind it.
Where does my data go?
Prompts and responses pass through Cloudflare’s infrastructure. We walk you through exactly what that means before anything is switched on, and if it is not acceptable we say so rather than sell it to you anyway. The logs are yours.
What happens if a provider goes down?
The gateway falls back to another provider and handles the retries, so an outage at one of them does not stop your application.
What do you need from me to get started?
The list of providers your applications use, and someone who can change an endpoint.
What happens when spending spikes?
Spend caps you set are enforced at the gateway, and rate limits hold per application. You are told before a budget is reached rather than after, and caching stops the same answer being paid for twice.
Is anything installed on our machines?
No. Nothing is installed on anyone’s machine, and the only change at your end is the endpoint your applications call.
Whose keys and whose logs are they?
Yours, both. Your providers stay your providers and your keys stay yours, and you own the logs. You can point your applications back at the providers whenever you want.
Do I need managed WordPress hosting or maintenance to buy this?
No. It runs as a standalone monthly service alongside whatever else you use. It is also part of AI Pro, if you would rather have the AI services on one invoice.
Ask Us Anything
We’d love to hear from you!