Everything You Need to Manage AI Traffic
A production-grade, Rust-powered AISIX gateway designed for performance, governance, and observability.
High Performance
Blazing Fast, Built with Rust
Native Rust data plane delivers sub-millisecond proxy overhead with minimal memory footprint. Handle millions of requests per second without breaking a sweat.
Unified Governance
One API, All Your LLMs
Manage all your LLM providers through a single, OpenAI-compatible API. Centralized configuration, authentication, and policy enforcement across your entire AI stack.
Intelligent Load Balancing
Multi-LLM Load Balancing
Dynamically distribute traffic across multiple LLM providers based on latency, cost, and availability. Weighted round-robin, least-connections, and custom strategies.
Precise Rate Limiting
Token & Request Rate Limiting
Fine-grained rate limiting by tokens, requests, or custom dimensions. Per-consumer, per-route, and cluster-wide policies to control costs and prevent abuse.
Security Guardrails
Enterprise-Grade Security
Protect your AI pipeline with prompt injection detection, content moderation, PII redaction, and comprehensive audit logging for regulatory compliance.
Rich Observability
Full-Stack Observability
Track every token, monitor latency distributions, and analyze traffic patterns in real-time. Native integration with Prometheus, Grafana, and ClickHouse.
Production-Grade,
Cloud-Native Architecture
Control Plane and Data Plane separation ensures high scalability, zero-downtime upgrades, and enterprise-level reliability.
Stateless Data Plane
Horizontally scalable with zero state β add or remove nodes instantly without data migration.
etcd-Based Config
Centralized configuration management with real-time propagation and hot-reload capabilities.
Separation of Concerns
Control Plane handles management; Data Plane handles traffic. Independent scaling and upgrades.
Deploy in Minutes, Not Weeks
OpenAI-compatible API. Production-ready from day one.
Enterprise-grade security built into every request
- TPM / RPM rate limiting β cap tokens-per-minute and requests-per-minute per key, per model
- Guardrails β block prompt injection, PII leakage, and toxic content before it reaches the model
- Per-key access control β restrict which models each API key can call
- Concurrency limits β prevent single consumers from monopolising capacity
100+ LLM Providers, One Gateway
Connect to any major LLM provider through a unified, OpenAI-compatible interface. No vendor lock-in, ever.
The Best Pricing Plans
Start free with our open-source core. Scale to enterprise when you're ready.
Open Source
Everything you need to get started with AISIX
- Connect 100+ LLM providers
- Load balancing + RPM/TPM limits
- Virtual key management
- OpenTelemetry, Prometheus, Jaeger
- Model guardrails
Enterprise
Cloud or Self-Hosted. For teams managing AI traffic at scale.
- Includes all open-source features
- Priority support + custom SLAs
- Budgets, teams, and RBAC
- Audit logs and compliance
- Complete enterprise features