This topic was automatically generated from Slack. You can find the original thread here.
!channel Hey y’all, we’ve been exploring some of the new capabilities that are unlocked with MCP and would love to know if you’ve all been using MCP already or if you’re thinking about specific use cases. More concretely, we’re working on exposing Pipedream triggers and actions as tools within MCP servers, so you could use them in any MCP client or SDK. We’re in prototyping and exploration mode on that front, and would love to just learn more about problems you’re trying to solve either within your core product or even as developers.
I use it all the time and see it as a complete game changer. Having built my own servers, setup my gtm clients on them, and experimented with it a lot - the biggest issue is around deployment. The mcp team is working on remote deployment and connection but since servers are currently setup to run locally I’ve had to find work around a dockerizing servers alongside the app + mcp client
in our case we’re hosting the server and exposing functionality via SSE — would you see yourself using our MCP servers to coordinate agents to call actions, deploy triggers, etc?
I could see myself using your MCP servers to abstract away a lot setup and add some autonomous workflow functionality for clients that could be cool.
And to your second point, that’s a good question and it is hard to say since it would depend on the use case but what comes to top of mind is I’ve been manually setting up client team members MCP walking them installing the required languages + packages to make it work on their machine - hosted would be better for those use cases
That sounds pretty cool, but I’d like to suggest that making Pi fully stable might be more important. It’s a fantastic resource, but it is very frustrating that it just stops generating at some point, without any explanation, warning, work-around, or information about when it will start working again.
I completely hear you, but we do have multiple teams working on products and so we are going to be working in parallel on new projects like this, in addition to improving the core product
I am using mcp servers in my local setup with cline. For example git, github, supabase (for updating tables, setting policies, …). And I wrote some custom ones (i.e. face generator). Not sure what I’d need for pipedream. However, I refrain using AI within pipedream, since tooling is always hard. What I usually do is use Flowise and just send an http request from pipedream. Having this one place would be awesome and much easier to handle. Also MCP won (imho). So betting on it feels less frightning
