guMCP is an open-source collection of Model Context Protocol (MCP) servers that can be run both remotely and locally. The project aims to create the largest collection of MCP servers with a unified backend, fostering a community around AI integrations and the future of AGI.
While many MCP server providers are closed source, and open-source alternatives typically only support local hosting through stdio, guMCP provides:
- A comprehensive collection of MCP servers that work both locally and remotely
- Support for both stdio and SSE (Server-Sent Events) transports
- A unified backend architecture for consistent implementation
- Full open-source access to encourage community contributions
🎬 RECOMMENDED: Watch our quick setup video before getting started!
This under 5-minute walkthrough will save you time and help you understand the key concepts.
- Python 3.11
- Git
- For Windows users: A bash-compatible shell is recommended (Git Bash, WSL, etc.)
-
Clone the repository:
git clone https://github.com/gumloop/guMCP.git cd guMCP -
Set up a virtual environment:
This isolates the project dependencies from your system Python installation.
# Create the virtual environment python -m venv venv # Activate it (choose the appropriate command for your OS) # On Unix/macOS: source venv/bin/activate # On Windows (Command Prompt): venvScriptsactivate # On Windows (PowerShell): .venvScriptsActivate.ps1 # On Windows (Git Bash): source venv/Scripts/activate
You’ll know your virtual environment is active when you see
(venv)at the beginning of your command prompt. -
Install dependencies:
# Install core dependencies pip install -r requirements.txt # Install development dependencies (for contributing) pip install -r requirements-dev.txt
-
Configure environment variables:
# Create a local environment file from the template cp .env.example .env # Open the .env file in your preferred text editor and update values as needed # If you're using VS Code: code .env
The
.envfile contains configuration for:- API keys for service integrations
- Authentication settings
- Server configuration options
- Development environment settings
An SSE server can be run
9 Comments
raunakchowdhuri
[dead]
rbehal1729
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.
vessenes
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.
yahoozoo
If you want to edit files on your local filesystem, do you have to have a MCP server running locally?
tuananh
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
compootr
> We're making one standardized mono project for all MCP servers
obligatory https://xkcd.com/927/
canadaduane
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?
billllll
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.
johnjungles
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!