ContextForge MCP Gateway - Secure Self-Hosted MCP Proxy

AWS Artificial Intelligence (AI)

ContextForge, the open source model context protocol gateway from IBM: federate many MCP servers behind one authenticated endpoint, with a registry and admin UI, secured from first boot by cloudimg.

Base
Hardened build
minimal ports, security patches applied at build time
Access
Unique credentials
generated on first boot, readable only by root
Verified
Boots working
services pass a health gate before release
Support
24/7, 365 days
by email and live chat, 24 hour response SLA

Overview

ContextForge MCP Gateway is an open source gateway, registry and proxy for the Model Context Protocol, the standard that lets AI assistants and agents call external tools. It sits in front of many separate MCP servers and presents them as one authenticated endpoint: it keeps a registry of the servers and the tools they expose, lets you compose virtual servers from a chosen set of tools, and normalises different MCP transports so a client connects once and reaches everything. An admin web interface manages the federated servers, tools and access, so teams get a single controlled front door to their MCP tool estate instead of wiring every assistant to every server by hand.

It suits platform and AI engineering teams who are running more than one MCP server and want central registration, discovery, authentication and auditing over the tools their agents can call, hosted inside their own cloud account.

Why the cloudimg image

cloudimg ships ContextForge secure by default with no surviving default login. On the first boot of every instance a unique administrator password, a JWT signing secret and an encryption secret are generated and written to a root only file, so no two deployments share a credential and the upstream default never governs a running instance. Because the gateway is a credential aggregation point for every server it federates, the gateway and its embedded database are bound to loopback behind a reverse proxy rather than exposed, and unauthenticated requests are refused. To prove the gateway genuinely works out of the box it ships one real reference MCP server on the appliance, pre federated at first boot, and the build verifies a real tool call travelling through the gateway to that server end to end. The base is fully patched with unattended security upgrades enabled, and every deployment is paired with a step by step guide and backed by 24/7 cloudimg support.

Common uses

  • One authenticated front door for many MCP servers
  • A central registry and admin UI for AI tool servers
  • Composing virtual MCP servers from selected tools

Key features

  • Prebuilt and secure in minutes: unlike a manual install that requires building a Python virtual environment, pinning dependencies, authoring a systemd unit, configuring a TLS reverse proxy and rotating every default secret by hand, this image completes every build and hardening step at image creation time. Launch an instance and reach a working, authenticated MCP gateway over HTTPS without editing a single configuration file.
  • Secure by default with unique credentials: upstream ships a known default admin and signing secret, so this image generates a cryptographically random administrator password, JWT signing secret and encryption secret on each instance's first boot, and mints a per-instance TLS certificate. No two deployments share a credential, the gateway and datastore bind only to the loopback interface, and only the nginx TLS reverse proxy is exposed.
  • Federate your whole MCP tool estate behind one endpoint: register Model Context Protocol servers, compose virtual servers from a curated set of tools, and hand your agents a single authenticated front door with a registry and admin UI. A bundled reference MCP server is federated out of the box so a real proxied tool call works the moment the instance boots, backed by 24/7 cloudimg technical support via email and chat.

See it running

Real screenshots taken while testing this image against its deployment guide.

ContextForge MCP Gateway - Secure Self-Hosted MCP Proxy screenshot 1 ContextForge MCP Gateway - Secure Self-Hosted MCP Proxy screenshot 2 ContextForge MCP Gateway - Secure Self-Hosted MCP Proxy screenshot 3 ContextForge MCP Gateway - Secure Self-Hosted MCP Proxy screenshot 4

Description

This is a repackaged open source software product wherein additional charges apply for cloudimg support services.

## Overview

ContextForge MCP Gateway is IBM's open source gateway, registry and proxy for the Model Context Protocol (MCP), the standard that lets AI assistants and agents call external tools. It sits in front of many separate MCP servers and presents them as one authenticated endpoint, with a registry of servers and their tools, virtual servers composed from a chosen set of tools, and an admin web interface. Rather than wiring each assistant to each server by hand, your agents connect once and reach your whole MCP tool estate through a single controlled front door. This AMI delivers the gateway fully installed and secured, exposed over TLS through a browser-based admin UI you run and control in your own AWS account.

## Why This Image vs a Manual Install

This image eliminates the Python virtual environment build, dependency pinning, systemd unit authoring, TLS reverse-proxy configuration, and credential hardening that a manual install requires. Upstream ships known default credentials and a known default signing secret, so a hand-rolled deployment is exposed until every secret is rotated by hand. This image generates a unique administrator password, JWT signing secret and encryption secret on each instance's first boot, binds the gateway and datastore to loopback, and fronts the UI with an nginx TLS reverse proxy. Because a gateway with nothing behind it cannot be demonstrated, the image also bundles one real reference MCP server, the official time server, federated automatically at first boot, so a working proxied tool call is available the moment the instance is up.

## Application Stack

ContextForge MCP Gateway is installed from the official upstream release (pip mcp-contextforge-gateway, pinned) into a Python virtual environment at /opt/mcpgateway, running as an unprivileged system user under a hardened systemd unit and bound to loopback 127.0.0.1:4444. An nginx reverse proxy terminates TLS on port 443 with a per-instance self-signed certificate and forwards to the gateway; port 80 redirects to 443. The registry, tools, virtual servers, users and tokens are held in a local SQLite datastore, so there is no external database to run. systemd starts the gateway, the bundled server and nginx on boot and restarts them on failure.

## Security Hardening

  • No default or shared credentials: the upstream default admin and signing secret never survive into a running instance; a random administrator password, JWT signing secret and encryption secret are generated on each instance's first boot and written only to a root-only file
  • Authentication required: the admin UI and gateway API refuse unauthenticated requests and any bearer token signed with a different secret
  • Network isolation: the gateway and bundled MCP server bind exclusively to loopback; only the nginx TLS reverse proxy is exposed
  • Per-instance TLS: each instance mints its own certificate on first boot, so no shared private key is baked into the image
  • Least privilege: the gateway runs as an unprivileged system user, and server-side request forgery protection stays enabled

Buyers requiring formal compliance documentation should contact cloudimg to discuss specific requirements.

## Concrete Use Cases

  • One front door for an agent's tools: A team building AI assistants points every assistant at a single authenticated gateway instead of wiring each one to each MCP server, and controls which tools are exposed from one admin console
  • Governed MCP tool estate: An organisation that wants its Model Context Protocol servers and the auth for them kept entirely inside its own AWS account, behind one controlled endpoint with a registry and per-instance credentials, rather than exposing each server directly
  • Virtual servers for a use case: A platform team composes a virtual MCP server from a curated set of tools drawn across several backend servers, and hands that single endpoint to a downstream agent

## AWS Integration

Deploy on any EC2 instance type. Front the gateway with your own DNS name and certificate by replacing the per-instance self-signed one. Restrict the security group so port 443 is reachable only from the agents and clients that need it, since the gateway aggregates credentials for every server it federates. Back up the SQLite datastore with EBS snapshots of the OS volume.

## Getting Started

Read your credential from the root-only file, open the admin UI over HTTPS, sign in as the administrator, and explore the bundled server and its federated tools. Register your own MCP servers, compose virtual servers, and mint API tokens for your agents.

## cloudimg Support

24/7 technical support by email and chat. Help with deployment, TLS termination with your own domain and certificate, federating your own MCP servers, and backup planning for your gateway datastore.

Related technologies

mcpmodel context protocolmcp gatewayai agentsai toolsllmgatewayregistryproxyself-hostedcontextforge