Sub-Millisecond Latency
Rust's zero-cost abstractions and Pingora's async architecture deliver routing decisions in microseconds. No garbage collector pauses, no runtime overhead.
Coming Soon — Join the Waitlist
The API Gateway
Built for Speed
Powered by Rust and Cloudflare's Pingora framework. Sub-millisecond routing, zero-copy proxying, and battle-tested reliability that handles millions of requests per second. A faster, lighter, and cheaper alternative to Kong, NGINX, Envoy, and Tyk.
10x
Faster than NGINX
50%
Less Memory
0ms
Config Reloads
No spam. Get notified when Tygress launches. Unsubscribe anytime.
tygress.yaml
gateway:
listen: "0.0.0.0:443"
tls: auto
routes:
- path: /api/v1/*
upstream: backend-cluster
rate_limit: 10000/s
circuit_breaker: trueFeatures
Everything You Need in a Gateway. Nothing You Don't.
Built on the same Pingora framework that powers Cloudflare's edge network serving trillions of requests daily. Everything Kong and NGINX charge for — included free in our open-source edition.
Memory Safe by Default
Rust's ownership model eliminates buffer overflows, use-after-free, and data races at compile time. Security without sacrificing speed.
Zero-Downtime Reloads
Hot-reload configuration without dropping a single connection. Pingora's graceful upgrade mechanism ensures seamless deployments every time.
Zero-Copy Proxying
Direct kernel-to-kernel data transfer using Pingora's optimized I/O paths. Proxy traffic with minimal CPU overhead and maximum throughput.
Full Protocol Support
Native HTTP/1.1, HTTP/2, gRPC, and WebSocket support built on Pingora's multi-protocol engine. Seamless protocol negotiation and upgrades.
Built-in Observability
Prometheus metrics, OpenTelemetry tracing, and structured logging out of the box. Full visibility into every request without external plugins.
Smart Rate Limiting
Distributed rate limiting with sliding window, token bucket, and leaky bucket algorithms. Protect your APIs without adding latency.
Dead Simple Config
Declarative YAML/TOML configuration that just works. Define routes, middleware, and upstreams in minutes, not hours. No Lua scripting required.
Extensible Plugin System
Write plugins in Rust or load WebAssembly modules for custom logic. Chain filters, transforms, and middleware with zero overhead.
AI Gateway
Not Just APIs. AI-Native.
Tygress is built for the AI era. Proxy, govern, and optimize LLM traffic with the same speed and reliability you expect for traditional APIs.
Unified Multi-Provider API
Single endpoint for OpenAI, Anthropic, Google, Mistral, and more. Switch providers or load-balance across them without changing a line of client code.
AI Governance & PII Redaction
Enforce usage policies, content filters, and automatically redact PII from prompts and responses before they reach upstream providers.
Semantic Caching
Cache LLM responses by meaning, not just exact match. Similar prompts hit the cache automatically — cut costs and latency by up to 90%.
Token-Aware Rate Limiting
Rate limit by token count, not just request count. Fairly throttle usage across teams and prevent a single heavy prompt from starving others.
MCP Gateway
Native Model Context Protocol support. Expose and manage MCP tool servers behind Tygress with auth, rate limiting, and observability built in.
Performance
Benchmarks Don't Lie. Tygress Is Faster.
Head-to-head against Kong, NGINX, Envoy, and Traefik on identical hardware (4 vCPU, 8 GB RAM). Rust + Pingora means no garbage collector pauses and zero-cost abstractions.
Tygress (Rust/Pingora)
p99: 0.12ms18 MB RSS
98K rps
Envoy (C++)
p99: 0.31ms52 MB RSS
72K rps
NGINX (C)
p99: 0.45ms38 MB RSS
65K rps
Kong (Lua/NGINX)
p99: 1.2ms128 MB RSS
34K rps
Traefik (Go)
p99: 1.8ms86 MB RSS
28K rps
Tyk (Go)
p99: 2.1ms94 MB RSS
24K rps
* Preliminary benchmarks. Final numbers will be published with the open-source release.
No Garbage Collector
Rust's ownership model means deterministic performance with no GC pauses — ever.
Async-First Architecture
Pingora's async runtime handles millions of concurrent connections on a single node.
Connection Pooling
Intelligent upstream connection reuse reduces TCP handshake overhead by up to 90%.
Kernel Bypass Ready
Optional io_uring and SO_REUSEPORT support for maximum throughput on modern Linux kernels.
Pricing
Powerful for Everyone. Priced to be Fair.
Start free with open source. Scale to enterprise. No hidden fees, no per-request charges, no surprises.
Open Source
Free Forever
$0
Full-featured gateway for developers and small teams.
- Core proxy & routing engine
- Rate limiting & circuit breaking
- Load balancing with health checks
- TLS termination
- HTTP/1.1, HTTP/2, gRPC, WebSocket
- YAML/TOML configuration
- Management dashboard
- Prometheus metrics
- Plugin system (Rust & WASM)
- Community support
Most Popular
Cloud
Fully Managed
$49/mo
Zero-ops gateway with a global edge network. Start in minutes.
- Everything in Open Source
- Managed control plane
- Global edge deployment
- Auto-scaling
- 99.99% uptime SLA
- Dashboard & analytics
- Team management
- Email & chat support
- Automatic updates
Cloud (Self-Managed)
Your Infrastructure
$99/mo
Run Tygress on your own cloud with our management plane.
- Everything in Cloud
- Deploy on your AWS/GCP/Azure
- Private networking support
- Custom domain & certificates
- Data residency controls
- Terraform & Pulumi providers
- SSO / SAML integration
- Priority support
- Audit logging
Enterprise
Unlimited Scale
Custom
For organizations that need the highest scale, security, and support.
- Everything in Self-Managed
- Unlimited nodes & clusters
- Multi-region federation
- Advanced RBAC & policies
- Custom plugin development
- Dedicated support engineer
- 24/7 phone & Slack support
- Custom SLA
- On-boarding & training
All prices are preliminary and may change before launch. Open source edition will always remain free.
Launching Soon
Be the First to Know
Join the waitlist to get early access, development updates, and exclusive launch-day pricing. We'll only email you when it matters.
Early access to beta
Launch-day discount
No spam, ever