One gateway for
every frontier model.
Supergate is a self-hosted LLM proxy that routes to OpenAI, Anthropic, Google, Cohere, and Mistral through a single OpenAI-compatible API — with semantic caching, multi-tenancy, BYOK, and per-key cost tracking built in.
// One client. Any model. Zero provider SDKs. const client = new OpenAI({ baseURL: 'https://your-gateway/v1', apiKey: 'gw_your_api_key', }) // Route to GPT-4o await client.chat.completions.create({ model: 'gpt-4o' }) // Route to Claude Opus 4.8 — same client await client.chat.completions.create({ model: 'claude-opus-4-8' })
Built different
Everything you need.
Nothing you don't.
Supergate covers the hard parts of production LLM infrastructure — so you can focus on building your product.
OpenAI-Compatible Endpoint
A single `/v1/chat/completions` endpoint handles all 5 providers. Change one string and your app routes to any model — no SDK changes, no adapter layers.
Semantic Caching via pgvector
Vector-similarity caching means near-duplicate prompts serve cached responses instantly. Cut your token spend by 30–50% with zero application changes.
Row-Level Tenant Isolation
Every query is scoped to a tenant ID enforced at the database level. Tenants cannot see each other's data, keys, or usage — ever.
Per-Key Cost Attribution
Every request logs model, provider, input/output tokens, latency, and dollar cost — attributed to the specific API key that made it.
Hybrid Key Management
Use our managed gateway keys for instant access, or store your own provider credentials with AES-256-GCM encryption. BYOK keys take precedence automatically.
Self-Hosted & Open Source
Deploy on your own infrastructure with a single `docker compose up`. MIT licensed. No vendor lock-in, no egress fees, no surprise bills.
Current model lineup
All frontier models.
One endpoint.
Models are kept up to date with provider deprecation schedules. Currently reflecting the June 2026 lineup.
The most capable frontier models for complex reasoning and generation.
Safety-first models with exceptional instruction following and analysis.
Multimodal-first models with massive context windows at low cost.
RAG-optimised models built for enterprise search and retrieval.
European open-weight models with strong multilingual performance.
How routing works
Supergate reads the model field in your request and routes to the correct provider automatically. BYOK keys take precedence over gateway keys per provider.
Flexible authentication
Managed keys or your own. Your choice.
Supergate supports a hybrid model — start with managed gateway keys, move to BYOK when you need compliance. Mix and match per provider.
Gateway Keys
Use Supergate's managed provider keys. Zero configuration — just generate a gateway key and start calling any model immediately. Ideal for prototypes and early-stage teams.
- No provider accounts needed
- Instant access to all 5 providers
- Unified billing through one gateway
- Great for rapid prototyping
Bring Your Own Key (BYOK)
Store your own provider credentials inside Supergate, encrypted with AES-256-GCM. Your keys are used directly — Supergate never proxies billing. Available per provider, per tenant.
- AES-256-GCM encrypted at rest
- Direct provider billing relationship
- Per-provider granularity
- Enterprise-Independent mode (BYOK-only tenants)
Transparent pricing
Start free. Scale when ready.
Or self-host entirely for free — all tiers are available on your own infrastructure.
Free
Get started instantly with managed gateway keys. No credit card required.
- All 5 providers via gateway keys
- 50K tokens / month
- Semantic caching included
- Usage dashboard
- 1 API key
- Community support
All plans include semantic caching, usage telemetry, and multi-tenancy. Self-hosting is always free.
Self-hosted in minutes
One command.
Fully running.
Supergate ships as a Turborepo monorepo with Docker Compose. Bring your own PostgreSQL and Redis, or let the compose file spin them up. Copy your .env, run the stack, seed the database, and you're live.