Scite MCP in ChatGPT

Hey there.

I'm Drew Barontini, VP of Product at Scite, and I want to show you the Scite MCP.

So on Scite AI MCP, we have all the instructions you need to get set up in different LLM tools that you want to use this in.

So depending on what you're using, you can follow the instructions there.

But I'm going to showcase ChatGPT today because it's really simple since we're now in the App Store.

So it makes it very simple for you to get this set up with your Scite subscription and to be using the, the Scite MCP inside of ChatGPT today.

So you just open it up in the ChatGPT store and you're going to go through the authentication flow.

So you can see I'm here in the apps area inside of ChatGPT.

There's a section in the sidebar you can get there.

I'm just going to search for Scite and we're going to see it come up here.

So you can see you have the information about the MCP.

I'm going to click on Connect.

I'm going to click Connect Scite.

This is going to open up a new tab in a moment and it's going to automatically sign me in.

I'm going to hit authorize, make sure my Scite account is connected.

So you need to have a Scite subscription in order for this to work.

Going to hit Start Chat and then you can see that it shows up down here on the bottom.

If it's not there, you can go into the plus more and click on Scite.

So you might have other MCPs in here that you want to use.

You can select which ones you want, but if you want to interface with Sites, you need to have that selected.

We're going to ask a question here.

So research question like, does trisomy increase the rate of chromosome segregation?

We're going to trigger this, request.

So this is going to.

ChatGPT is going to use that Scite MCP connection that's currently in place, and it's going to make a request out to Scite.

It's going to search the literature and it's going to return a response and show that to you inside of the standard channel ChatGPT experience.

But leveraging that unique data set that we have at Scite.

So this is gonna go run behind the scenes and it will print that out when it has the response for you.

Okay, so you can see now we have a response coming back and it's going to give an answer, but you can go through and ask whatever Follow up questions you want.

This is looking at the literature, this is calling out specific references inside of the response.

So you can see it actually searched across the literature and you can see all of the response coming back as well, which is really cool because you can look at that to get more of that validity and validation behind the scenes because you can see where it's coming back from.

So it's asking some specific questions here.

I can say what are the supporting citations?

So I'm going to ask a question that's specific to what we do at cite with our citation graph and with smart citations because we're actually looking at our, all of the citations inside of an individual body of work, individual literature, and we're looking at how that relates to other parts of the literature that's being referenced.

So this is something that can be asked and you can get more answers around some of that validation with those different citations across the literature.

So you can see it's going to come back and it's going to tell you here are the supporting ones.

It's going to call out very specific references with specific DOIs that you could look up directly if you need to it's showing you specific context here.

And if I open up this response as well, you can see it's looking at a quite large response set.

It's pulling back lots of data here and it's going to use that and parse that down really quickly so you can see it called out a couple of the supporting citations with that.

And you continue to follow up, follow up with this, ask more questions and you can do this across a range of tools that you already use, like ChatGPT, like Claude on the Web or desktop Claude code and other tools as well.

We're really excited for a lot of the possibilities here and would love to get feedback.

So if you have anything, please send feedback our way so we can keep making this better for you and your research workflows.

Thanks.