Drop-in for the code you have
Point the OpenAI or Anthropic SDK at the gateway and swap the key. Streaming, tool calls, JSON output, images and PDFs work as they did.
Anthropic · OpenAI · xAI — one endpoint
Give every app its own AI key. Keep the limits in your hands.
A1 API Gateway sits between your software and the model vendors. Issue a virtual key for each app, teammate or customer, lock it to IP addresses, rate limits and a hard spending cap, and see what every request costs.
- OpenAI-compatible API
- Native Anthropic Messages API
- Caps enforced before the call
from openai import OpenAI
client = OpenAI(
base_url="https://api.a1apigateway.com/v1",
api_key="sk-gw-…", # a virtual key
)
reply = client.chat.completions.create(
model="anthropic/claude-sonnet-5",
messages=[{"role": "user", "content": "Hello"}],
)A raw vendor key is an open tab
Hand one Anthropic or OpenAI key to every app and a single leak exposes the whole account: no spending limit, no idea who used what, and rotating it breaks everyone at once.
Unlimited spend
A leaked or looping key keeps billing until someone notices.
No attribution
The vendor invoice shows one total, not which app or customer ran it up.
Painful rotation
Replacing a shared key means redeploying everything that uses it.
Guard rails
Three limits on every key, checked before the vendor is called
Each virtual key carries its own rules. A request that would break one is refused at the gateway, so it never reaches Anthropic, OpenAI or xAI and never costs anything.
- Where it can be used from. An IP allowlist, allow-anywhere, or pinned to the first address that uses it.
- How fast. Requests per minute, tokens per minute and concurrent requests.
- How much. Lifetime, monthly, weekly, daily and hourly dollar budgets, plus a monthly ceiling for the whole organisation.
activeinvoice-parser-prod
sk-gw-4k2p…9xq- Where
- 203.0.113.0/24 only
- How fast
- 120 requests · 200k tokens a minute
- How much
- $250 a month, $20 a day
This month$154.80 of $250
Everything between your code and the model vendors
Metering, routing, isolation and reconciliation in one place, so each team and customer gets exactly the access you intend.
Every request priced
Each call is priced in US dollars and attributed to a key, a model and a vendor key, so you can see exactly where the money goes.
Your vendor keys, or our credit
Bring your own Anthropic, OpenAI or xAI keys and be billed by them directly, or run on prepaid credit that can never go negative.
Isolated by design
Every organisation gets its own console on its own subdomain and its own database file. There is no shared table to leak across.
Reconciled every day
Gateway figures are checked against Anthropic's and OpenAI's own usage reports daily, and differences are flagged.
Built for teams
Owners, admins and viewers, mandatory two-factor sign-in, alerts by email or signed webhook, and an audit log of every change.
Live in an afternoon
- 1
Create your organisation
You get a private console at
your-name.a1apigateway.com, protected by two-factor sign-in. - 2
Connect a way to pay for tokens
Add your own vendor keys, which are tested and encrypted before they are stored, or arrange prepaid credit with us.
- 3
Issue keys and change one URL
Create a key per app, set its limits, and point your SDK at
https://api.a1apigateway.com.
The models you already use
Claude, GPT and Grok through one OpenAI-compatible endpoint, with Anthropic's own Messages API for teams that want it. Organisations choose which models are switched on, and aliases let you change a model without a deploy.
- Anthropic Claude
- OpenAI GPT
- xAI Grok
- Streaming
- Tool calling
- Structured JSON
- Images
- PDF documents
- Prompt caching
- Extended thinking
- Token counting
Pricing that fits how you use AI
Tell us which vendors you use and roughly how much, and we will put together a quote.