- Home
- LLM gateways
- Envoy AI Gateway
Envoy AI Gateway
What it is
Envoy AI Gateway, now called Agent Router, is an open-source control plane for managing AI and agent traffic through Envoy. It provides an OpenAI-compatible API for hosted and self-hosted models and MCP servers, with integrations including OpenAI, Azure OpenAI, Google Gemini, AWS Bedrock, Mistral and Anthropic. Traffic features include model virtualization, provider failover, token quotas and usage-based rate limits. Its MCP gateway offers server multiplexing, tool routing, OAuth authentication and observability. InferencePool support adds routing and load balancing for inference endpoints. Provider credentials can be managed centrally, including short-lived credential flows for AWS Bedrock, Azure OpenAI and GCP Vertex AI. The local CLI runs on Linux and macOS; Kubernetes production deployments use Envoy Gateway and must meet listed compatibility requirements. The open-source project costs nothing and uses the Apache 2.0 license. The project says its code, maintainers and APIs remain the same after the rename, and directs users to Discord and GitHub issues for support.
For and against
For
- OpenAI-compatible API covers models and MCP servers.
- Includes provider failover, quotas and usage-based rate limiting.
- Central credential management keeps keys out of application workloads.
- Apache 2.0 licensed and free.
- Supports local Linux and macOS CLI use.
Against
- Local CLI support is limited to Linux and macOS.
- Kubernetes deployment requires compatible dependencies.
- Support is through Discord and GitHub issues.
The verdict
Agent Router brings model access, traffic controls and MCP routing under one open-source control plane. Teams should check the deployment requirements against their environment before adopting it.
Who it suits
It suits teams that share models or tools and want central credentials, quotas, usage attribution or provider failover. It also fits teams prepared to operate a local proxy or a Kubernetes deployment with compatible dependencies.
Plans and prices
Open-source project
Self-managed; local CLI supports Linux and macOS · Kubernetes deployments require compatible dependencies
Read 9 Oct 2026 · theagentrouter.ai
On the record
- Credential security
- Upstream credentials can be centrally managed; supported providers include short-lived credential flows for AWS Bedrock, Azure OpenAI, and GCP Vertex AI.theagentrouter.ai · read 9 Oct 2026
- Deployment
- The CLI can run locally on Linux and macOS, and production deployment uses Kubernetes with Envoy Gateway.theagentrouter.ai · read 9 Oct 2026
- Deployment options
- The project can run locally as a standalone proxy or deploy to Kubernetes; the site also links to hosted services offered by other providers.theagentrouter.ai · read 9 Oct 2026
- Inference routing
- InferencePool support provides intelligent routing and load balancing for inference endpoints.theagentrouter.ai · read 9 Oct 2026
- Intended users
- The documentation identifies teams sharing models or tools, seeking centralized credentials, quotas, usage attribution, or failover as likely users.theagentrouter.ai · read 9 Oct 2026
- Kubernetes compatibility
- The compatibility matrix lists Agent Router v1.2.x with Kubernetes v1.32 or later, Envoy Gateway v1.9.2 or later, and Gateway API v1.6.1.theagentrouter.ai · read 9 Oct 2026
- License
- The project uses the Apache 2.0 license.theagentrouter.ai · read 9 Oct 2026
- Local platform support
- The `aigw run` CLI currently supports Linux and macOS.theagentrouter.ai · read 9 Oct 2026
- MCP gateway
- The MCP gateway aggregates tools from multiple servers and supports tool routing, security policies, and observability.theagentrouter.ai · read 9 Oct 2026
- Observability
- The CLI supports OpenTelemetry tracing and can record request and response details; the documentation describes configuration to hide input and output content.theagentrouter.ai · read 9 Oct 2026
- Provider integrations
- The site lists integrations including Anthropic, AWS Bedrock, Azure OpenAI, Cohere, DeepInfra, DeepSeek, Google Gemini, Grok, Groq, Mistral, OpenAI, Together AI, and Vertex AI.theagentrouter.ai · read 9 Oct 2026
- Purpose
- The open-source project routes and manages generative AI traffic between applications and AI services, with rate limiting and policy control.theagentrouter.ai · read 9 Oct 2026
- Renamed project
- Envoy AI Gateway is now Agent Router, an Agentic AI Foundation project; the site says the code, maintainers, and APIs remain the same.theagentrouter.ai · read 9 Oct 2026
- Security
- The project describes upstream authentication for securing connections to AI services and says provider credentials can be kept out of application workloads.theagentrouter.ai · read 9 Oct 2026
- Support
- The project directs users to its community Discord and GitHub issues for help.theagentrouter.ai · read 9 Oct 2026
- Traffic controls
- Capabilities include model virtualization, provider failover, token quotas, and usage-based rate limiting.theagentrouter.ai · read 9 Oct 2026
- Unified API
- It provides an OpenAI-compatible API for routing requests across model providers and lets applications switch models by changing the model name.theagentrouter.ai · read 9 Oct 2026
