MCPJungle
MCPJungle is an open-source, self-hosted Model Context Protocol (MCP) gateway for developers and teams. It lets users register multiple MCP servers once and expose their tools, prompts, and resources through one endpoint, with optional centralized access control and observability for shared deployments.
At a Glance
- Individual developers
- Engineering and platform teams
- Organizations operating internal AI-agent infrastructure
- Enterprise teams requiring authenticated, governed MCP access
AI Tools by MCPJungle
(1)
MCPJungle
Self Hosted MCP Server Gateway
Discussions
No discussions yet
Be the first to start a discussion about MCPJungle
Latest News
MCPJungle 0.4.7 released with vulnerable Go-module upgrades, including x/net, to Go 1.26.9.
MCPJungle 0.4.6 released with configurable server bind host support via --host / MCPJUNGLE_BIND_HOST.
MCPJungle 0.4.5 released with configurable SQLite database-file paths.
MCPJungle 0.4.4 released; the documentation identifies the Dashboard UI as a beta local/development management interface.
Products & Services
A self-hosted gateway and registry that registers remote Streamable HTTP and local STDIO MCP servers and exposes them through a unified /mcp endpoint.
Standalone command-line client and binary for starting the server, registering servers, managing tools, prompts, resources, groups, and enterprise identities. Distributed through GitHub Releases and Homebrew.
Beta local/development web interface for inspecting and managing MCP servers, tools, tool groups, prompts, resources, and system information.
Container deployment artifacts for local Docker Compose use and shared enterprise deployments, including standard and STDIO-enabled images.
Market Position
MCPJungle positions itself as a free, open-source, self-hosted MCP gateway and registry rather than a hosted SaaS service. Its differentiators are a single endpoint for multiple MCP servers, local-first deployment, CLI and Docker distribution, Tool Groups for context control, and an enterprise mode with authentication, ACLs, PostgreSQL, and metrics. Relevant alternatives in the broader MCP gateway category include Docker MCP Gateway, Microsoft MCP Gateway, IBM MCP Gateway, and Lunar.dev MCPX.
Leadership
Founders
Raghav Dua
Site Reliability Engineer at Booking.com; GitHub identifies him as the person building MCPJungle under the @mcpjungle organization. His GitHub handle is duaraghav8.
Executive Team
Raghav Dua
Founder and maintainer
Site Reliability Engineer at Booking.com and the GitHub account identified as building MCPJungle.
Founding Story
Raghav Dua said he had been working on MCPJungle for a while before open-sourcing it. The project was started to solve the operational problem of deploying multiple AI agents and MCP servers: without a gateway, every client needs separate configuration, tools are scattered, and access control is duplicated or absent. The initial vision was a single, self-hosted control point for registering MCP servers and serving them behind one endpoint.
Business Model
Revenue Model
MCPJungle is currently free to use for both individuals and enterprises. It is distributed as open-source software under the Mozilla Public License 2.0, with self-hosted binaries, Docker/GHCR images, and Homebrew distribution. The public documentation does not describe a paid subscription, API-usage charge, or support plan.
Pricing Tiers
Available for both individual development-mode use and enterprise-mode shared deployments.
Target Markets
- Individual developers
- Engineering and platform teams
- Organizations operating internal AI-agent infrastructure
- Enterprise teams requiring authenticated, governed MCP access
- Developers consolidating personal MCP configurations into one local endpoint
- Teams operating shared MCP infrastructure
- Centralized discovery and governance of MCP tools
- Reducing context overload by exposing tool subsets through Tool Groups
- Connecting multiple AI clients and agents to a common MCP environment
- Production or internal-platform deployments using enterprise mode, PostgreSQL, authentication, ACLs, and observability