agentgateway vs Higress
agentgateway
agentgateway.dev
Higress
higress.ioSide by side
| Figure | agentgateway | Higress |
|---|---|---|
| Price from | Free | Free |
| Free plan | Yes | Yes |
| Free trial | — | — |
| Open source | Yes | Yes |
| Web | Yes | Yes |
| Windows | Yes | Yes |
| macOS | Yes | Yes |
| Linux | Yes | Yes |
| API | Yes | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 1 | 5 |
| Spec score · Kubernetes Ingress Controllers | 9.1 (#1) | 8.6 (#4) |
| Deployment | The project provides standalone binary, Docker and Kubernetes deployment paths. | A Docker all-in-one deployment exposes the console on port 8001, HTTP gateway on 8080 and HTTPS gateway on 8443. |
| Licence | The open-source project is licensed under Apache-2.0. | The main Higress project is licensed under Apache License 2.0, while bundled components can carry separate licences. |
| Limit: operational complexity | A reviewer said the Kubernetes controller and policy model add too much machinery for a single service with one model provider and a fixed tool list. | The same evaluation judged Higress better suited to Kubernetes platform teams than developers expecting a conventional plain go mod installation. |
| Observability | Built-in observability exports OpenTelemetry metrics, logs and distributed traces. | The AI statistics plugin records token counts, first-token latency, total request latency and logs across gateway, route, service and model dimensions. |
| Security | Listed security features include JWT and OIDC, API keys, authorization policies, TLS and mTLS with certificate rotation, and rate limiting. | Listed protections include prompt-attack detection, sensitive-content identification, data privacy protection, and output-content filtering. |
| Support | The enterprise page lists partner-provided commercial support, including a Solo.io offer with 24×7 coverage, upstream collaboration, and best-practice guidance. | The community edition includes community support and documentation; the dedicated commercial edition lists ticket, DingTalk group, and offline support. |
| What it does | Agentgateway is a Rust data plane that connects, secures, routes and observes LLM, MCP, A2A and ordinary service traffic. | Higress is an open-source, AI-native cloud API gateway built on Istio and Envoy. |
Plans
Higress
- Alibaba Cloud AI Gateway Serverless StandardPriced on request
- Higress CommunityFree
- Higress Community EditionFree
- Alibaba Cloud AI Gateway Serverless EnterprisePriced on request
- Feitian Exclusive EditionPriced on request
In their own words
agentgateway
Agentgateway is a free, open-source HTTP and gRPC gateway for routing, securing, observing, and governing service, LLM, MCP, and agent-to-agent traffic. Its LLM gateway offers an OpenAI-compatible API for providers including OpenAI, Anthropic, Bedrock, Gemini, Vertex, and OSS Llama runs. Self-hosted inference routing supports vLLM, TGI, and Triton, with choices guided by latency, cost, and model. For MCP it supports server discovery, scoped tool calls, and audit trails; A2A routing connects runtimes such as LangChain, CrewAI, and ADK. Security features include JWT and OIDC, API keys, authorization policies, TLS and mTLS with certificate rotation, and rate limiting. Guardrails inspect MCP calls and responses for prompt injection and unsafe payloads, while prompt guards support checks and redaction. Observability includes OpenTelemetry, Prometheus token-usage histograms, Jaeger traces, and request logs with latency, token counts, and realized USD cost. It can run as a standalone binary, in Docker, or on Kubernetes. Commercial support is listed through a partner.
Higress
Higress is an open-source API gateway for governing AI services and cloud-native traffic, built on Istio and Envoy. Its AI Gateway provides centralized access to model APIs and MCP servers, with unified routing, protocol conversion, load balancing, and fallback. Higress says it supports more than 100 mainstream large language models. It can convert HTTP to MCP and proxy native MCP services. Traffic controls include token-based quotas and rate limits, API key pool rotation, and consumer authentication. Caching can reuse inference results precisely or semantically to reduce token use and latency for similar or repeated questions. Listed security measures cover prompt attacks, sensitive content, data privacy, and output filtering. The wider API gateway combines traffic, microservice, and security gateway functions with service exposure and API lifecycle management. It supports HTTP, HTTP2, HTTPS, GRPC, GRPCS, TCP, and Dubbo, along with Kubernetes Ingress and Gateway API. Custom Wasm plugins can be written in Go, Rust, or JavaScript, with sandbox isolation and hot updates. The self-hosted Community Edition is free; the site also lists a managed Alibaba Cloud plan billed by actual usage, starting at ¥0.