Every developer gets
an AI budget.
Automatically.
Your company pays for AI, but nobody knows who's using it — or how to stop one person from spending it all. FoundryGate hands out the keys, watches the meter, and stops overspending the moment it starts.
One page tells you everything.
This is the kind of view an admin gets: every developer, how much of their monthly AI budget they've used, and who's asking for more. Budgets come in a few set sizes — a standard one, a bigger one for heavy users, and unlimited for the few who need it. The names below are made up — the dashboard is real.
You can put a money figure next to those numbers too. You type in what your AI usage costs — Microsoft's bill arrives as one lump sum, so nobody but you knows how to split it — and the dashboard multiplies it out per person. It's an estimate, and it's labelled one everywhere it appears. Until you fill your prices in, no money figures show at all.
| Developer | Monthly budget | Used | Status |
|---|---|---|---|
| Priya Raghavanpriya.raghavan@contoso.com | 20,000,000 tokens | 62% | Active |
| Marcus Webbmarcus.webb@contoso.com | 5,000,000 tokens | 38% | Active |
| Sofia Delgado-Ruizsofia.delgado@contoso.com | 20,000,000 tokens | 91% | Near limit |
| Tomasz Kowalskitomasz.kowalski@contoso.com | 5,000,000 tokens | 100% | Blocked until Oct 1 |
| Jenna Okaforjenna.okafor@contoso.com | Unlimited | — | Unlimited |
| Diego Fuentesdiego.fuentes@contoso.com | 5,000,000 tokens | 64% | Asked for more |
| Erik Lindqvisterik.lindqvist@contoso.com | 5,000,000 tokens | 12% | Active |
| Aisha Rahmanaisha.rahman@contoso.com | 20,000,000 tokens | 77% | Active |
Three answers. That's the whole system.
From the developer's side there's nothing to install and nothing to learn. They point their usual coding tool at your gateway, and every request comes back one of three ways.
The answer comes back, and with it a note saying what it cost and how much budget is left.
Going too fast this minute. The gate says how long to wait, and coding tools wait on their own.
The monthly budget is gone. Requests stop the same second — and the message says when it comes back.
Not a mockup. A running gateway.
These are real responses from a live FoundryGate gateway, captured during testing on September 1, 2026. This is exactly what a developer's tools see.
HTTP/1.1 200 OK x-fg-remaining-quota: 97668 x-fg-remaining-tpm: 19984 x-fg-tokens-consumed: 16
HTTP/1.1 429 Too Many Requests Retry-After: 50
HTTP/1.1 403 Forbidden { "statusCode": 403, "message": "Token quota is exceeded. Try again in 29 days, 9 hours, 50 minutes, and 5 seconds." }
$ codex exec "Reply with exactly: pong" pong tokens used: 9,139
Captured from a live FoundryGate gateway on 2026-09-01. Header names and timings are unedited.
Live evidence Every test run against a real deployed gateway is written up in thevalidation/ folder of the repository — one file per run, with the
raw responses.
Start wherever you like.
The docs go from plain English to full detail. Take the first card if you want the idea, the second if you want the picture, the rest when you're ready to build.
Your Azure. Your models. Your rules.
FoundryGate is free and open source. It runs entirely inside your own Azure account — nothing leaves it, and no third party ever sees your usage or your data.