@ericdallo welcome!

Few things I'd like to share about MCPJungle:

1. It's a Registry + Gateway Proxy
   You can register and track all your MCP servers in mcpjungle.
   All your MCP clients (claude, cursor, your ai agents) just need to connect to a single URL - the mcpjungle gateway - to access all the MCP tools.

2. It allows you to control which clients can access which mcp servers using ACLs

3. It is currently self-hosted
   You run mcpjungle on your local or your own server. Either directly or inside Docker.
   We're evaluating providing a SaaS offering but don't have any definite plans right now.

4. Things that we'll release in the next 1 month

   • Support for STDIO MCP servers
   • a GUI with search features
   • Improved compatibility with Claude code & desktop and VSCode

If you can tell me your use case, I can help provide even more clarity and even prioritise some features.
I'm happy to hop on a call to give you a Demo as well.

As for Natoma, although I'm not an authority on it, here's my take:

1. It gives you access to all the publicly available MCP servers like github, jira, etc. Mcpjungle currently cannot do that out-of-the-box, but you can deploy the mcp and then just register it in mcpjungle.
   But I'm not sure if Natoma allows you to integrate your own private MCP servers and make them available via the Gateway. Mcpjungle firstly focuses on your private MCPs (or public MCPs that you'd like to deploy in your private environment).

2. It provides a SaaS offering, so you don't host anything yourself. If you're comfortable with sharing your data & mcp interactions with a third-party, then this is easier.

3. It provides metrics and audit logs about MCP interactions. Mcpjungle plans to provide this in the near future but doesn't have it yet.

If you have a lot of enterprise needs like identity, security & audit, compliance, infra-as-code, etc today, perhaps Natoma works for you at the expense of being a lot more complex as a product.

Mcpjungle aims to be super simple & lightweight, but as of today it is not suitable for large enterprise deployments that have a lot of strict requirements.

Happy to hear your thoughts!