Start the AI Gateway in One Command.
OpenAI & Anthropic compatible. One API for every LLM provider.
npx -y iauroraBuilt for raw speed & efficiency
Aurora Gateway is engineered in Go and validated with external side-by-side load tests against Bifrost using the same host, payload, and mocked upstream.
External Side-by-Side Results
| Metric | Aurora | Bifrost | Winner |
|---|---|---|---|
| Successful requests | 192,026 | 174,471 | Aurora |
| Success rate | 64.01% | 58.16% | Aurora |
| Throughput | 3,130.18/s | 2,888.09/s | Aurora |
| Mean latency | 833.14ms | 912.47ms | Aurora |
| P50 latency | 709.16ms | 729.60ms | Aurora |
| P99 latency | 2.155s | 2.107s | Bifrost |
| Peak memory | 59.47MB | 113.11MB | Aurora |
5000 requested RPS for 60 seconds using the same external load generator, same mocked OpenAI-compatible upstream, 25 warmup requests, and a 256-request prewarm.
Tested Under Real Gateway Pressure
Aurora was measured side-by-side against Bifrost with the same host, traffic profile, model payload, and OpenAI-compatible upstream.
Reproduction details live in the benchmark docs, so the homepage stays focused on the performance story.
Core OSS Pillars, Enterprise Gates
Practical governance, routing, and observability controls for teams moving AI traffic into managed production environments.
Governance
Manage OSS-safe API keys, usage, workflows, and audit logs from the admin plane; add tenants, users, roles, and budgets with Enterprise capabilities.
- Auth & Audit: Master keys, managed keys, and auditable request history in OSS; OIDC and RBAC in Enterprise.
- Usage First: Usage tracking works in OSS; forecasts, resets, and scoped spend controls require budgets.
- Admin API: Configure providers, pools, auth keys, workflows, and guardrails; tenant routes are gated.
Routing & Caching
Route requests through provider pools, fallback chains, passthrough APIs, and exact or semantic response caching.
- Provider Pools: Tune upstream selection and refresh runtime configuration.
- Fallback Chains: Keep requests moving when a configured upstream degrades.
- Built-in Cache: Inspect exact and semantic cache behavior from the dashboard.
Security Controls
Combine authentication, scoped keys, audit logging, guardrail rules, and private deployment practices for safer AI traffic.
- Guardrail Rules: Apply configured input/output policy through the gateway pipeline.
- PII Controls: Use redaction-oriented guardrails where sensitive content must be reduced.
- Deployment Control: Keep the gateway inside your own network and storage boundary.
One gateway surface for apps and agents.
Point OpenAI- and Anthropic-compatible clients, provider-native passthrough clients, and server automation at Aurora to centralize routing, managed keys, audit logs, usage analytics, caching, and admin controls.
from openai import OpenAI client = OpenAI( api_key="change-me", base_url="http://localhost:8080/v1", ) response = client.chat.completions.create( model="<model id @@kw@@from /v1/models>"@@end@@, messages=[{"role": "user", "content": "Hello"}], )
Deploy in Seconds
Start with npx, ship with Docker or native binaries, and keep Enterprise delivery private.
npx -y iauroraCore controls stay on when Enterprise features stay off.
Model Catalog
List, route, and manage models across configured providers, aliases, and provider-specific deployments.
Usage Analytics
Track requests, tokens, cost, models, providers, user paths, and time ranges without enabling Enterprise budgets.
Fallback & Pools
Use fallback chains and provider pools to route traffic across configured upstreams.
Response Caching
Use exact and semantic response caching with cache controls, overview metrics, and request debugging.
Native Passthrough
Forward provider-native requests through /p/{provider}/... when full upstream API flexibility is required.
Admin Dashboard
Manage providers, pools, auth keys, workflows, guardrails, audit logs, usage, and OSS-safe settings.
Scoped API Keys
Create managed keys with user paths, rate limits, expiry, provider/model controls, and usage statistics.
OpenAI- & Anthropic-Format APIs
Expose chat, responses, embeddings, models, files, batches, and messages for OpenAI- and Anthropic-compatible clients.
Observability
Inspect usage, cost, audit logs, console streams, cache behavior, provider health, and Prometheus metrics.
Open Source
A Go-based gateway designed for local, self-hosted, and private AI infrastructure.
Trusted by engineering teams worldwide.
Aurora gives teams one place to inspect usage, tune caching, and manage provider routing instead of scattering that logic across applications.
Sarah Chen
CTO, Synthwave Labs
Aurora's OpenAI- and Anthropic-compatible surface lets teams keep client code small while centralizing auth keys, audit logs, and provider configuration.
Marcus Rodriguez
Head of Engineering, DataForge AI
The dashboard brings usage, budgets, cache behavior, audit logs, and provider health into one operational view.
Priya Kapoor
VP of Infrastructure, Pulse Analytics
Self-hosting Aurora keeps the gateway, storage, and provider credentials inside the team's own operating boundary.
James Whitfield
Security Architect, NexGen Financial