What it is

Agentgateway is an open source proxy written in Rust and licensed under Apache-2.0. It is a next generation agentic proxy for AI agents and MCP servers, and it operates inside the Developer Tools, API gateway, Kubernetes, service mesh, and AI gateway ecosystem. The project is built on the MCP and A2A protocols.

The concrete problem it addresses is the need to secure, observe, and govern traffic between agents, tools, models, and external services. Agent-to-LLM, agent-to-tool, and agent-to-agent communication can be spread across frameworks and environments, which makes routing, authentication, policy, and observability harder to manage. Agentgateway presents itself as a drop-in proxy for those connections.

Key capabilities

• LLM Gateway routes traffic to OpenAI, Anthropic, Gemini, and Bedrock through a unified OpenAI-compatible API with budget controls, prompt enrichment, load balancing, and failover.
• MCP Gateway connects LLMs to tools and external data sources through MCP, with tool federation, stdio, HTTP, SSE, Streamable HTTP transports, OpenAPI integration, and OAuth authentication.
• A2A Gateway enables secure agent-to-agent communication using A2A, with capability discovery, modality negotiation, and task collaboration.
• Inference Routing sends requests to self-hosted models using Kubernetes Inference Gateway extensions, based on GPU utilization, KV cache, LoRA adapters, and queue depth.
• Guardrails provide content filtering through regex, OpenAI moderation, AWS Bedrock Guardrails, Google Model Armor, and custom webhooks.
• Security and observability include JWT, API keys, OAuth, RBAC with CEL policy engine, rate limiting, TLS, and OpenTelemetry metrics, logs, and tracing.
• The project includes a built-in UI for exploring agent-to-agent or agent-to-tool connections.

Who uses it and how

• Agent developers route model traffic through the LLM Gateway and apply budget and spend controls.
• Platform teams expose tools and external data sources to LLMs through the MCP Gateway, using stdio, HTTP, SSE, or Streamable HTTP transports.
• Kubernetes operators deploy agentgateway with the built-in controller and Gateway API to manage agentic traffic.
• Teams without a full Kubernetes controller use standalone deployment driven by flat YAML configuration.
• Developers and operators use the built-in UI to inspect agent-to-agent or agent-to-tool connections.

Getting started

Typical start paths are the standalone quickstart for flat YAML configuration and the Kubernetes quickstart for deployment with the built-in controller and Gateway API.

When to use it — and when not to

Use agentgateway when you want an open source Rust proxy for MCP, A2A, LLM routing, guardrails, and Kubernetes Gateway API workloads. Avoid it when you need a long production track record, because the repository is about one year old, has 258 open issues, and lists no contributor count. The supplied material does not mention paid products it replaces, Docker image names, package managers, or required database, storage, or SMTP services.

The <strong>first complete</strong> connectivity solution for Agentic AI.

  </a>
</td>