Skip to content
sugusGateway
OpenAI-compatible

One key. Every model.
Priced per request.

sugusGateway is an OpenAI-compatible AI gateway over a single upstream. Point an existing SDK at it, get every model behind one key, and see the exact cost of every single request — streamed, as it happens.

base_url = /v1

~/sugus
curl /v1/chat/completions \
  -H "Authorization: Bearer $SUGUS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4.5",
    "stream": true,
    "messages": [
      {"role": "user", "content": "Explain CRDTs in three sentences."}
    ]
  }'
Spend this weekEvery request, priced in micro-USD.
$0.00$0.50$1.00$1.50$2.00MonTueWedThuFriSatSun
Spend this week: 7 points from Mon to Sun. Total $5.02, highest day Sun at $1.12.
PeriodSpend
Mon$0.42
Tue$0.58
Wed$0.51
Thu$0.77
Fri$0.93
Sat$0.69
Sun$1.12
$1.12
  • OpenAI-compatible wire format
  • No SDK changes required
  • Integer micro-USD accounting

Three things the upstream API does not do

The gateway is additive: every field your model providers already emit is passed through untouched. These are the extras.

  • Unified access

    Every model the gateway serves behind one OpenAI-compatible endpoint. Point an existing SDK at sugusGateway and the request body is forwarded byte-for-byte — nothing is lost in translation.

  • Per-request cost

    Every request is priced in integer micro-USD and returns its own cost in the response. Logs carry the upstream price next to yours, so margin is a number, not a guess.

  • Streaming that measures

    Server-sent events pass through untouched, and a running cost estimate rides along in each chunk. Watch spend accrue live, then get the exact figure on the final usage frame.

Bring your own key, see the real number

The playground streams a response and shows the cost of that response, token by token, as it arrives.