Skip to main content
Glama

MCP Gateway

Every MCP call from your agents flows through Glama – so you know exactly what your AI is doing, and you decide what it's allowed to do.

MCP Client
Glama MCP Gateway
MCP Server

Every call logged · Every tool gated · Every credential managed

What is an MCP Gateway?

An MCP Gateway is a reverse proxy that sits between AI clients and MCP servers. It appears as an MCP server to the client while acting as an MCP client to the upstream server. Without a gateway, every agent manages its own connections to every server – a tangled point-to-point mess that breaks at enterprise scale. The Glama MCP Gateway consolidates all of that into one control plane where every call is visible, auditable, and revocable.

Route any MCP endpoint

The Gateway is source-agnostic. It fronts every kind of MCP endpoint with the same logging, access control, credential management, and analytics:

  • Open-source MCP servers hosted on GlamaDeploy any server from the registry to Glama's infrastructure in one click. The Gateway fronts it automatically.
  • Hosted MCP connectorsPre-indexed remote services like Linear, Stripe, and PostHog. Connect through the Gateway for centralized auth, call logging, and per-tool access control.
  • Any remote MCP serverAny MCP endpoint you already run or want to use – bring your own credentials, point the Gateway at the URL, and route the traffic through your control plane.

How to connect

The Gateway handles two flows depending on what you're routing. Pick the path that matches your server, paste the URL into any MCP-compatible client, and you're done.

Open-source servers hosted on Glama

Deploy any server from the registry to Glama and we automatically put the Gateway in front of it. No proxy setup, no manual wiring – the endpoint you get is the Gateway.

  1. Pick a server from the registry.
  2. Click Deploy on Glama – Glama builds and hosts it on dedicated infrastructure.
  3. A Gateway endpoint is generated automatically when the deployment is healthy.
  4. Paste the URL into Claude, Cursor, ChatGPT, or VS Code.
https://glama.ai/endpoints/<your-connection-profile>/mcp

Every call through this URL is logged, access-controlled, and credential-managed automatically.

Browse MCP servers →

Hosted MCP connectors

Remote MCP services like Linear, Stripe, and PostHog. You have two ways to connect – the Gateway is opt-in.

Option 1 – Direct

Use the connector's native endpoint. Fastest path, but you manage credentials and have no logging or access control.

https://mcp.<service>.com/mcp
Option 2 – Through the Glama Gateway

Create a connection profile for the connector you want to use. You get a dedicated URL backed by call logging, per-tool access control, managed OAuth, and usage analytics.

https://glama.ai/endpoints/<your-connection-profile>/mcp
Browse connectors to route →

What you get

Three things a direct MCP connection cannot give you. Everything else the Gateway does – credential hygiene, SLA protection, rate limiting – is cumulative to these.

Unified access

  • Single URL for every MCP endpoint you connect
  • Managed OAuth 2.1 with automatic refresh

Full observability

  • Every JSON-RPC message logged with full payloads
  • Audit log export for your SIEM

Tool-level control

  • Per-tool access control with approval flows
  • Rewrite upstream tool definitions

Gateway vs. proxy

An MCP proxy is a thin pass-through: a JSON-RPC request comes in, gets forwarded upstream, and a response goes back. Nothing else. A gateway keeps the proxy and layers access control, encrypted OAuth storage with automatic token refresh, session lifecycle tracking, aggregated usage analytics, and full call logging on top. Use a proxy for basic routing against one upstream server. Use the Glama MCP Gateway when you are running MCP at scale and need governance.

Put the Glama MCP Gateway in front of your agents

Plus hosted MCP deployments on dedicated infrastructure, an OpenAI-compatible LLM gateway, and cost attribution labels.