Control and optimize every AI call.
A single unified API to give your team access to 100+ models across providers with built-in routing, observability and organizational management.
100+ models. One endpoint.
1 / 5
openai/gpt-5.2
GPT · frontier model
meta/llama-4-maverick
Llama · open weights
elevenlabs/scribe_v1
ElevenLabs · speech to text
anthropic/claude-4.5-haiku
Claude · fast reasoning
deepseek/deepseek-v4-pro
DeepSeek · reasoning model
google/gemini-3.1-flash
Gemini · fast multimodal
mistral/ministral-8b
Mistral · edge model
openai/gpt-5.2
GPT · frontier model
meta/llama-4-maverick
Llama · open weights
elevenlabs/scribe_v1
ElevenLabs · speech to text
anthropic/claude-4.5-haiku
Claude · fast reasoning
deepseek/deepseek-v4-pro
DeepSeek · reasoning model
google/gemini-3.1-flash
Gemini · fast multimodal
mistral/ministral-8b
Mistral · edge model
anthropic/claude-opus-4.8
Claude · reasoning model
google/gemini-3-pro-preview
Gemini · frontier model
elevenlabs/eleven_multilingual_v2
ElevenLabs · text to speech
openai/gpt-5.2-mini
GPT · fast model
deepseek/deepseek-coder-v4
DeepSeek · code model
meta/llama-4-scout
Llama · open weights
mistral/mistral-large-3
Mistral · frontier model
anthropic/claude-opus-4.8
Claude · reasoning model
google/gemini-3-pro-preview
Gemini · frontier model
elevenlabs/eleven_multilingual_v2
ElevenLabs · text to speech
openai/gpt-5.2-mini
GPT · fast model
deepseek/deepseek-coder-v4
DeepSeek · code model
meta/llama-4-scout
Llama · open weights
mistral/mistral-large-3
Mistral · frontier model
google/gemini-3-flash-lite
Gemini · cheapest tier
openai/gpt-5.2-audio
GPT · speech to speech
meta/llama-3.3-70b
Llama · open weights
anthropic/claude-sonnet-4.6
Claude · balanced model
mistral/codestral-2
Mistral · code model
elevenlabs/eleven_flash_v3
ElevenLabs · low latency
deepseek/deepseek-chat-v4
DeepSeek · general model
google/gemini-3-flash-lite
Gemini · cheapest tier
openai/gpt-5.2-audio
GPT · speech to speech
meta/llama-3.3-70b
Llama · open weights
anthropic/claude-sonnet-4.6
Claude · balanced model
mistral/codestral-2
Mistral · code model
elevenlabs/eleven_flash_v3
ElevenLabs · low latency
deepseek/deepseek-chat-v4
DeepSeek · general model
Production AI needs better inference infrastructure.
Hicap is that infrastructure: two products behind one endpoint.
POST /v1/chat/completions
Run & manage
Hicap Control Plane
- Every request logged with team, app, and model.
- Budgets and rate caps enforced at the gateway.
- Keys, roles, and audit history in one place.
Token price optimization
Hicap Inference
- Requests routed to the cheapest model that clears your bar.
- Reserved and spot capacity priced below pay-as-you-go.
- Savings land on the invoice, line by line.
Observability and control for every token.
Everything your AI does: logged, tagged, budgeted, and access-controlled — in one place.
Request log · requested → served
● live
| Time | Requested | Served | Route | Latency | Status |
|---|---|---|---|---|---|
| 14:02:11 | gpt-5.5 | gpt-5.5 | direct | 320ms | 200 |
| 14:02:11 | claude-opus-4-8 | claude-opus-4-8 | direct | 608ms | 200 |
| 14:02:12 | claude-opus-4-8 | → gemini-3.1-pro | failover · 429 | 441ms | 200 |
| 14:02:12 | gemini-3.1-flash | gemini-3.1-flash | direct | 184ms | 200 |
| 14:02:13 | auto | → mistral-large-3 | cost-optimized | 296ms | 200 |
Every reroute is logged: the model you asked for, the model that answered, and why.
Know which model actually answered.
Every request writes one log line: the model you asked for, the model that answered, and what it cost. Enable deeper analytics and tracing to see usage over the last 7 days, per team, app, or key.
- One log line per request
- Requested vs served model
- Usage analytics over the last 7 days
Model Usage
Cost, volume & tokens by model across all apps
Top Models by Credits
Spend leaderboard
#1 eleven_multilingual_v2
29,161.00
#2 eleven_v3
27,708.00
#3 scribe_v1
3,046.00
#4 scribe_v2
2,658.00
#5 claude-opus-4.7
1,191.00
Every request tagged by feature, team, app, and model.
Add tags — feature, team, app — to each API request, and Hicap records them on every log line and rolls them into spend attribution automatically. Set them once in your client; there's no separate tracking pipeline to maintain.
- Tag requests by feature, team, and app
- Every request logged with team, app, model, feature
- Drill down without extra setup
Set spend limits per team, user, or key.
Give every team, app, or key its own budget. When a limit is reached, access can be restricted — no more guessing what's allowed.
- Per-team, per-user, per-key budgets
- Configurable limits and quotas
- Restrict access when limits are reached
Application
Tooling-Devactive
Created Jun 10, 2026 · basic · demo@hicap.aiPrimary Key
RevealConnection Analytics
9.1M
Total Tokens
4.0K
Requests
40,102.00
Credits Used
7.7M
Input Tokens
2.6M
Output Tokens
962.8K
Cached Tokens
Daily Token Usage
All Time
Top API Keys Leaderboard
Highest-spending API keys across all applications
#1 Tooling-Dev
sub-mq88spd4
$401 4.0K
#2 Tooling-Prod
sub-mq88soii
$289 1.0K
Every key and model, under one policy.
Give each team scoped keys and a list of approved models, and every request they make is checked against it. You always know which apps run, on which models, with whose keys.
- Connection-scoped, rotatable keys
- Model allow-lists per team
- Full audit trail, exportable
Infrastructure built for inference. One gateway to every model.
Point your OpenAI-compatible client at api.hicap.ai and every request rides a distributed gateway: load-balanced across providers and regions, health-tracked, and rerouted mid-flight when a provider fails. Your app stays up.
99.9%
Uptime, enforced by auto-failover
429 · 5xx · timeout → reroute
0% markup
Pay provider list price, nothing added on top
transparent billing
1 endpoint
Distributed gateway, OpenAI-compatible
api.hicap.ai
P50 · P95 · P99
Latency percentiles per model and route — optimized through routing
every provider
Routing strategies
Every token, cheaper.
Lock in the throughput you already use at up to 25% below pay-as-you-go, and let spikes overflow to on-demand. Whatever you reserve but don't use sells on the spot market, and the revenue returns to you.
Reserved capacity
MonitoringListed38k tokens on the spot market
$0.62/tok
Instant integration.
Point your existing OpenAI-compatible client at api.hicap.ai and every model is live.
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 }'
Works with your stack.
One OpenAI-compatible endpoint in front of every major provider. Point your existing tools at Hicap and keep everything else exactly as it is. No framework lock-in.
Drop-in access
Integrations
Model providers
100+ compatibilities
Secure by default. Sovereign from day one.
Private networking, no data retention, tenant isolation, and role-based access — out of the box.
Private networking
Direct model inference that bypasses the public internet
No data retention
We do not store your prompts, responses, or model outputs. Your data never leaves the request path.
Data residency
Tenant isolation and residency controls per deployment
Role-based access
Permissions scoped down to the connection and key
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.