Just a couple of things that I think makes GuMCP different than other options out there:
- Having a unified way to run each server via stdio and sse. Basically all existing providers either let you connect through their hosted sse service or let you clone and run stdio locally. Rarely is there a very simple way to do either/or, especially when it comes to hosting your own sse server out-of-the-box. Existing options like Zapier's MCP are closed source, so it's not so simple.
- Setting a flexible and generic framework for integration auth with MCP servers. Currently, methods of authentication vary from server to server. We provide a unified mechanism of auth for every single server that is importantly generic enough for you to be able to host it on your own. We do this through having a base 'AuthClient' that is used across each of our servers, which supports any arbitrary implementation. For local use, we provide a LocalAuthClient that sets a standard for integrations, OAuth and otherwise.
- We also have a GumloopAuthClient that uses your existing credentials for Gumloop, which will then exact similarly to how Zapier or Composio do their MCP servers, with auth taken care of easily. Difference is you have the flexibility to also host it on your own with your own AuthClient, or run locally with stdio and local auth as well.
- Ready-to-deploy sse through one URL. If you want to host on your own, the repository out-of-the box can be deployed with ALL sse servers hosted under a single URL.
- For open source contributions, your server will be deployed and hosted instantly. No need to fiddle with devops to get a SSE server available remotely for your implementation. This makes it way simpler for less technical folks (or just those who don't want to deal with infra) to make their ideas a reality.
We did this because of a painpoint I experienced as an engineer having to deal with crummy mcp setup, lack of support you have no idea how hard it is to set up SSE, deal with API keys and scope issues, and then to find things like the tool that you want isn’t even coded yet.
And so one of the areas we found it to be useful was to do the soft communications with tools like Jira linear slack - updating stakeholders and all that friction that engineers hate doing. Some other areas people use a lot of tools with sequential thinking
Mostly for workflows that I like:
* start a PR with a summary of what I just did
* slack or comment to linear/Jira with a summary of what I pushed
* pull this issue from sentry and fix it
* Find a bug a create a linear issue to fix it
* pull this linear issue and do a first pass
* pull in this Notion doc with a PRD then create an API reference for it based on this code
* Postgres or MySQL schemas for rapid model development
Everyone seems to go for the hype but ease of use, practical pragmatic developer workflows, and high quality polished mcp servers are what we’re focused on
Lmk what you think!
Currently working on a desktop app so everything runs on a native Mac app!
very similar to my goals but i went different way.
I chose wasm for package format. so everyone can choose the programming language they are familiar with. caveat: rust still has best support for wasm. everything else is hit or miss.
I chose OCI registry for package publishing because that's what everyone already has in their infrastructure. or just use dockerhub/GHCR.
Anyone else hosting MCP servers in a generally available way? Seems like that's the end goal given the network transports (especially the one in new spec revision).
Seems like everything I find requires running locally or hosting your own instance.
We did this because of a painpoint I experienced as an engineer having to deal with crummy mcp setup, lack of support you have no idea how hard it is to set up SSE, deal with API keys and scope issues, and then to find things like the tool that you want isn’t even coded yet.
And so one of the areas we found it to be useful was to do the soft communications with tools like Jira linear slack - updating stakeholders and all that friction that engineers hate doing. Some other areas people use a lot of tools with sequential thinking
Mostly for workflows that I like:
* start a PR with a summary of what I just did
* slack or comment to linear/Jira with a summary of what I pushed
* pull this issue from sentry and fix it
* Find a bug a create a linear issue to fix it
* pull this linear issue and do a first pass
* pull in this Notion doc with a PRD then create an API reference for it based on this code
* Postgres or MySQL schemas for rapid model development
Everyone seems to go for the hype but ease of use, practical pragmatic developer workflows, and high quality polished mcp servers are what we’re focused on
I like this! MCP is still very new, and a new paradigm, and super powerful. Psyched to see it integrated in gumloop. Is the intent to have a sort of app store of MCP servers eventually? Feels like it would be useful.
I'm particularly interested in typescript MCP servers, as they integrate well with the web infra ecosystem I work with. Do you foresee a typescript collection of guMCP servers as well?
Yeah, I guess my question was: that assumes the MCP server is running on the same machine as your files? Otherwise, if it's a remote MCP server then you would have to be developing on that remote machine? That may be an obvious "yes" so I just wonder why there are so many posts around hosting MCP servers via Cloudfare or what have you if they can't even edit your files on your machine.
Thanks HN! (cofounder here)
Just a couple of things that I think makes GuMCP different than other options out there:
- Having a unified way to run each server via stdio and sse. Basically all existing providers either let you connect through their hosted sse service or let you clone and run stdio locally. Rarely is there a very simple way to do either/or, especially when it comes to hosting your own sse server out-of-the-box. Existing options like Zapier's MCP are closed source, so it's not so simple.
- Setting a flexible and generic framework for integration auth with MCP servers. Currently, methods of authentication vary from server to server. We provide a unified mechanism of auth for every single server that is importantly generic enough for you to be able to host it on your own. We do this through having a base 'AuthClient' that is used across each of our servers, which supports any arbitrary implementation. For local use, we provide a LocalAuthClient that sets a standard for integrations, OAuth and otherwise.
- We also have a GumloopAuthClient that uses your existing credentials for Gumloop, which will then exact similarly to how Zapier or Composio do their MCP servers, with auth taken care of easily. Difference is you have the flexibility to also host it on your own with your own AuthClient, or run locally with stdio and local auth as well.
- Ready-to-deploy sse through one URL. If you want to host on your own, the repository out-of-the box can be deployed with ALL sse servers hosted under a single URL.
- For open source contributions, your server will be deployed and hosted instantly. No need to fiddle with devops to get a SSE server available remotely for your implementation. This makes it way simpler for less technical folks (or just those who don't want to deal with infra) to make their ideas a reality.
Interesting, thanks!
Would something like https://github.com/OpenAdaptAI/OmniMCP make sense to include here?
Oh this is really interesting!
I think this is a good complement for use WITH guMCP rather than having it in the core repo itself, definitely will be playing around with this.
If you want to try out mcp (model context protocol) with little to no setup:
I built https://skeet.build/mcp where anyone can try out mcp for cursor and dev tools.
We did this because of a painpoint I experienced as an engineer having to deal with crummy mcp setup, lack of support you have no idea how hard it is to set up SSE, deal with API keys and scope issues, and then to find things like the tool that you want isn’t even coded yet.
And so one of the areas we found it to be useful was to do the soft communications with tools like Jira linear slack - updating stakeholders and all that friction that engineers hate doing. Some other areas people use a lot of tools with sequential thinking
Mostly for workflows that I like:
* start a PR with a summary of what I just did * slack or comment to linear/Jira with a summary of what I pushed * pull this issue from sentry and fix it * Find a bug a create a linear issue to fix it * pull this linear issue and do a first pass * pull in this Notion doc with a PRD then create an API reference for it based on this code * Postgres or MySQL schemas for rapid model development
Everyone seems to go for the hype but ease of use, practical pragmatic developer workflows, and high quality polished mcp servers are what we’re focused on
Lmk what you think!
Currently working on a desktop app so everything runs on a native Mac app!
hey! we'll build the same thing but just open source it and make it free :)
very similar to my goals but i went different way.
I chose wasm for package format. so everyone can choose the programming language they are familiar with. caveat: rust still has best support for wasm. everything else is hit or miss.
I chose OCI registry for package publishing because that's what everyone already has in their infrastructure. or just use dockerhub/GHCR.
https://github.com/tuananh/hyper-mcp
Anyone else hosting MCP servers in a generally available way? Seems like that's the end goal given the network transports (especially the one in new spec revision).
Seems like everything I find requires running locally or hosting your own instance.
If you want to try out mcp (model context protocol) with little to no setup hosted for you:
I built https://skeet.build/mcp where anyone can try out mcp for cursor and dev tools.
We did this because of a painpoint I experienced as an engineer having to deal with crummy mcp setup, lack of support you have no idea how hard it is to set up SSE, deal with API keys and scope issues, and then to find things like the tool that you want isn’t even coded yet.
And so one of the areas we found it to be useful was to do the soft communications with tools like Jira linear slack - updating stakeholders and all that friction that engineers hate doing. Some other areas people use a lot of tools with sequential thinking
Mostly for workflows that I like:
* start a PR with a summary of what I just did * slack or comment to linear/Jira with a summary of what I pushed * pull this issue from sentry and fix it * Find a bug a create a linear issue to fix it * pull this linear issue and do a first pass * pull in this Notion doc with a PRD then create an API reference for it based on this code * Postgres or MySQL schemas for rapid model development
Everyone seems to go for the hype but ease of use, practical pragmatic developer workflows, and high quality polished mcp servers are what we’re focused on
Lmk what you think!
if you go to gumloop.com/mcp all these servers are hosted for you with auth handled via your free gumloop account.
I like this! MCP is still very new, and a new paradigm, and super powerful. Psyched to see it integrated in gumloop. Is the intent to have a sort of app store of MCP servers eventually? Feels like it would be useful.
I'm particularly interested in typescript MCP servers, as they integrate well with the web infra ecosystem I work with. Do you foresee a typescript collection of guMCP servers as well?
I'm building a registry of typescript servers here: https://github.com/wegotdocs/open-mcp
Definitely at some point in the near future! started with python for now just because we had to pick a lane but we'll be porting them all over one day
If you want to edit files on your local filesystem, do you have to have a MCP server running locally?
The simplest way is to have a command that the MCP server can hit to interface with the filesystem.
This is the simplest way I know of to get started with it.
https://modelcontextprotocol.io/quickstart/user
Yeah, I guess my question was: that assumes the MCP server is running on the same machine as your files? Otherwise, if it's a remote MCP server then you would have to be developing on that remote machine? That may be an obvious "yes" so I just wonder why there are so many posts around hosting MCP servers via Cloudfare or what have you if they can't even edit your files on your machine.
> We're making one standardized mono project for all MCP servers
obligatory https://xkcd.com/927/
actually made me laugh. will be using this in the future
[dead]
[dead]