Agentgateway MCP
agentgateway
A proxy for agentic traffic: routes and secures connections to MCP servers, LLMs, and other agents over A2A
- 6,144 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Sits in the path of all agent traffic to MCP servers, LLMs and other agents
- Configures authentication and access policies; a misconfiguration can open up unintended access
Install
- Works in
Manual install
Deploy agentgateway using the standalone quickstart at agentgateway.dev/docs for a single service, or the Helm charts under controller/install/helm for Kubernetes. There is no single one-line command in the README; follow the documentation section for your scenario.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Sits in the path of all agent traffic to MCP servers, LLMs and other agents
- Configures authentication and access policies; a misconfiguration can open up unintended access
This is third-party code. Review the repository files before installing.
What it does
Agentgateway is a Rust proxy that sits between agents and whatever they talk to: MCP servers, LLM providers, and other agents over the A2A protocol. It federates several MCP servers behind one entry point, supports stdio, HTTP, SSE and Streamable HTTP transports, and adds OAuth authentication plus OpenAPI integration. For LLM traffic it exposes one OpenAI-compatible interface to different providers with load balancing and spend controls. It also has a guardrails layer with content filtering and CEL-based access policies, plus OpenTelemetry metrics and tracing. It deploys as a standalone process or as a Kubernetes controller via the Gateway API.
Who it is for. For devops engineers and teams who need to manage agent access to multiple MCP servers and LLMs from one place.
Good fit when
- Your team has several MCP servers and agents and needs one authenticated entry point
- You need to restrict which tools agents can reach based on roles and policies
- You need routing and spend control across requests to different LLM providers
Not a fit when
- You have a single local MCP server for personal use and don't need a proxy in front of it
- You have no infrastructure team to deploy and maintain a proxy layer
Example request
Set up agentgateway so it federates my MCP servers behind one entry point with OAuthMCP
- Transport
- stdio, http, sse
- Authentication
- OAuth
Limitations
Requires its own deployment and configuration, either as a standalone service or via Helm on Kubernetes. The README doesn't give a single one-line local run command; separate documentation pages cover the standalone and Kubernetes setups.
How to disable. Stop the agentgateway process or pod, or remove the Helm release in the cluster (helm uninstall for the agentgateway-crds and agentgateway-standalone charts).
FAQ
Is this an MCP server or a proxy in front of MCP servers?
A proxy. Agentgateway doesn't implement tools itself; it federates and secures access to existing MCP servers, LLMs and other agents.
Does it work outside Kubernetes?
Yes, there's a standalone mode driven by flat YAML config, separate from the Kubernetes controller with Gateway API support.
Discussion
No questions or reviews yet.
Related
MCP serversA self-hosted gateway routes internal MCP tools, authenticates users and records requests for later review.
MCP Gateway
mcp-gateway
A reverse proxy and lifecycle management for MCP servers on Kubernetes and Azure from Microsoft: routing, authorization, roles. For platform teams
A self-hosted MCP and A2A gateway with identity provider integration, access control and observability for enterprise agents
A self-hosted gateway between AI apps and models, MCP tools and A2A agents: policy, budgets, routing and an audit trail
MCP Gateway Registry
mcp-gateway-registry
One registry and gateway for MCP servers, agents and skills: shared login, OAuth, audit, OpenTelemetry. Community project for Docker and cloud
An IBM gateway and registry that federates MCP servers, A2A agents and REST or gRPC APIs behind one endpoint with auth and observability
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Agentgateway (this card) | yes | OAuth | 6,144 | |
| AIHelms MCP Gateway | no | API key | 1,035 | An organizational MCP gateway that gives employees personal keys and tracks tool calls together with their costs |
| Shepherd Model Gateway | no | 717 | The CLI runs a model gateway that routes requests across backends and executes MCP tools under configured policies. |
README in short
The README calls agentgateway the first complete connectivity solution for agentic AI and lists four layers: an LLM gateway with one OpenAI-compatible API, an MCP gateway with tool federation across transports, an A2A gateway for agent-to-agent communication, and inference routing for self-hosted models based on GPU metrics. It also describes guardrails with content filtering across several providers and a security layer with JWT, API keys, OAuth, CEL-based RBAC and OpenTelemetry telemetry. Setup points to separate documentation pages for standalone and Kubernetes scenarios, and the repository ships ready Helm charts.