LLM Gateway
One API for every model.
One key and an OpenAI-compatible endpoint put every major provider behind the client you already run. Hicap routes each request to the right model and fails over automatically when a provider wobbles.
Just want to integrate? See how it works.
Access
One key and one OpenAI-compatible endpoint put every major provider behind the client you already run.
Route
Each request goes to the cheapest model that clears your quality bar — frontier only when it is actually needed.
Stay up
Automatic failover across providers keeps answers flowing when one rate-limits, degrades, or goes down.
The reality
Every team built its own AI stack.
Marketing, support, engineering, finance: each with their own models, clouds, keys, and bills. It all works. And nobody can say what ran, where, or what it cost.
Hicap Platform
Now they all run on one.
Hicap replaces this with one endpoint, one invoice, one set of rules.
One unified API layerLower costs
Higher reliability
Nothing unaccounted
One key. Total control.
One key. Every major provider.
Point your existing OpenAI-compatible client at api.hicap.ai and the whole catalog is live. Change a model string to change providers; nothing else in your code moves.
1curl https://api.hicap.ai/v1/chat/completions \2 -H "api-key: $HICAP_API_KEY" \3 -H "Content-Type: application/json" \4 -d '{5 "model": "gpt-5.4",6 "messages": [7 { "role": "user", "content": "Hello" }8 ]9 }'
One key
A single credential opens every major provider. No per-provider contracts, quotas, or key rotation calendars.
Drop-in compatible
The OpenAI-compatible clients you already run work unchanged: chat, streaming, tool calls, and embeddings.
One bill
Usage across every provider lands on a single invoice, priced and itemized the same way everywhere.
The same work, off the frontier bill.
Visibility is only half the story. Because Hicap sits on the request path, it does not just report the bill, it changes it: routine work moves to cheaper models and the ledger shows the price each request avoided.
Routing ledger
Cost / 1M tokens
classify
-$7.30
gpt-5.5 → llama-4-405b
$9.40 → $2.10
summarize
-$8.30
claude-opus-4-8 → claude-sonnet-4-6
$12.60 → $4.30
extract
-$4.90
gemini-3.1-pro → mistral-large-3
$7.80 → $2.90
reason
kept
claude-opus-4-8 → claude-opus-4-8
$12.60 → $12.60
billed $21.90 of $42.40 list
saved $20.50
Route down by default
Routine work runs on models that cost a fraction of the frontier and clear the same quality bar.
Frontier where it counts
Hard reasoning stays on the strongest models. The policy decides per request, not per project.
Savings you can audit
Every routing decision is logged with the price it avoided, so the number on this ledger is checkable.
Smarter routing means better performance for less.
Every request is automatically routed to the best model for its needs, whether that's lower cost, faster speed, or higher quality. Your top models are only used when they're truly needed.
incoming request
↑
Built to stay up when a provider goes down.
Scale is not just more tokens per second. It is throughput that holds while providers rate-limit, degrade, and recover underneath you.
Automatic failover
Requests retry across providers on 429s, 5xx errors, and timeouts. Your app sees an answer, not an incident.
Provider health tracking
Latency and error rates are watched per provider, and traffic drains away from a degrading one before it fails.
Reserved-first throughput
Reserved capacity absorbs your baseline at a fixed price. On-demand catches the burst above it.
One endpoint through it all
Failover, bursting, and provider swaps happen behind api.hicap.ai. Application code never learns about the weather.
The one-line migration
Bring order to your AI stack in one line of code.
Change a base URL and every model, every provider, and every dollar runs through one place.