Deepgram Saga: The Voice OS for Developers

--- type: transcript date: 2025-07-11 youtube_id: uQjxmt1CWso --- # Transcript: Deepgram Saga: The Voice OS for Developers In this video, I'm going to be showing you Deepgram Graham Saga, the voice OS for developers. And what Deepgram Graham Saga allows you to do is to create your own Siri based on the tools that you actually want to leverage. This works with tools that we already use, whether it's cursor, MCP, or Slack. I'm going to show you how you can set this up. I'm going to show you how you can include your own MCP servers and also demonstrate the speed as well as accuracy on what you can leverage this for to increase your productivity as well as workflows. Let's get started. One quick aside, this is going to be free for the first thousand people that download this, go ahead and check it out within the description of the video if you're interested in trying this out. Once you download this and install it, what you can do with this is if you're within something like cursor or cloud code, I can go within the prompt window and I can command quotation mark. I can say something like I want to build out a nexation that leverages superbase for the back end and I also want to leverage better off for authentication. And then finally for emails I want to leverage Amazon SCES. And what I'll do is I'll command quotation mark again. And what we'll see within here is we see this cursor prompt. If I just expand this a little bit here, we'll see that it wrote out this enhanced prompt. What I can do with this is if I command enter, it will actually go ahead and replace that text within the prompt window. This is a much better initial prompt than what I initially said. And where this can be helpful is not just within something like cursor or cloud code. You can leverage this within chat GBT or even for something like writing an email or writing something like a Slack message. There are a ton of really neat features built within the platform, but one of my favorites is this MCP support. MCP or model context protocol if you're not familiar. Basically, what it allows us to have is a standardized format for how we can communicate different tools. And where that can be helpful is all that we need to actually set up different tools is to configure the server just like this. And within that object, we'll have a number of different services that we want to incorporate. Say within here, we want to leverage the file system or an external service. We can paste all of that within here similar to how you would within something like cursor or windsurf or cloud code. If you have used MCP before, this is going to be very familiar. Now, in terms of probably the easiest way that I've found in incorporating MCPs, including accessing services that do need authentication, things like Gmail or Google or Linear, you can go over to Zapier and they'll manage all of that. If you've used Zapier before for Zaps and different workflows, it's going to feel very familiar. But what's great with this is you can go and explore their thousands of different tools. They have over 7,000 applications. You're going to be able to combine a number of different services and create these dynamic workflows with whatever you want to leverage. And the beauty with MCP is it's going to be able to know when to invoke which tools to actually accomplish the task that you're asking of it. I have a handful of tools set up within here, but I really encourage you try this out with the applications that you actually use within here. I have Linear, I have Google Docs, as well as Gmail. And to set these up, it's as simple as going through. You can click to log into Google calendar. You can select the type of action that you want to have. Say if I want to have an MCP action that will find a particular event within my calendar. You could imagine I go over to saga and I ask to find that event and it will be able to perform that functionality to accomplish that task and then return that response back to me with natural language just like the Siri that we wish we had. Right? So what you can do at any point is you can trigger saga with option spacebar on Mac. That's how you're going to be able to open it up. Alternatively, you can have this larger view if you prefer that. And now to demonstrate some of the MCP capabilities and where this is really helpful. This is what I mean where it's like the Siri we wish we had where we can incorporate the tools that we actually want to use. And what's great with Zapier is you can go and you can include all of the things that you want. They already have the infrastructure in place to manage your Google Docs or whatever you have in mind that you want to leverage whether it's Linear or Slack. any of that is going to be all within there. Next, what I want to demonstrate, so on my desktop, I have this file called RD.txt. So, what I'm going to do is I'm just going to close this out just so you see that everything is working. Now, what I gave Saga access to is my desktop as well as those MCP integrations that we set up that I showed you within Zapier. What you can do with Saga at any point to open up the assistant, you can click option space. Alternatively, you can open it up in the full view like I have here. Now, just to demonstrate this, I'm going to say, "What files do you have access to?" I have access to the desktop and documents directories on your computer. Okay. What files do I have on my desktop right now? Your desktop contains files like prd.ext and directories. That prd one, can you tell me what's within that? The prd file outlines a five-step development plan for a next.js web application covering project setup. Perfect. That's the one. Now, I want you to take the contents of that document and let's add all of those as relevant issues within linear and add in all the relevant issues from your PRD to your linear tracker including configuration tasks, UI component, data integration, routing, and testing. All right, and here we go. So, here are all of the issues that it added within linear. And if I just go over to the text document that I had here, so it broke out that PRD into relevant issues here. Within here, I see core UI components and layout. I have the issue for that. I also see within here we have page structure and routing. All of that's within here. If I click through within the issues, we do also have a description for all of them. So, you can begin to see as you get creative with this for the particular workflows that you leverage day-to-day. It is really neat to experiment with having voice as that new medium to communicate. You can both speak faster as well as accomplish things by being able to integrate MCP for a number of different applications that you use. So you can leverage this within cursor or cloud code or actually have it for the different workflows and project management software that you have, let alone being able to leverage it within something like Gmail or Google Docs. Just to demonstrate this, I'm going to say create a Google doc that says, "Hello, my name is Developers Digest." This is an example of me leveraging Saga of actually creating a document. Here we can see that it's processing the request. Exactly. Your Google doc is ready with the specified content. Now if I go over to Google Docs here is exactly what I asked for. So just one quick last demonstration within this is I'm going to say create an email to developers digest.ai@gmail.com and I wanted to read that hey I got really great exciting news. I want to extend a job offer to you. Will you be able to meet over the next couple days? I drafted an email to developers digest.ai offering a job and asking to meet in the next few days. And then here we go. We have the email. We have the subject line that it was able to determine that all through the power of MCP as well as how you can leverage these language models. And if you're not as familiar with MCP or function calling or tool calling, the really cool thing with the underlying technology is it's going to be able to detect from natural language all of the relevant pieces of what it needs to pass into the relevant tool if a tool is detected to be used. For instance, for Gmail, we have different arguments, for instance, that we'd have to pass in such as the email, the subject line, as well as the body. And based on what I said with just one prompt of natural language, it was able to break down exactly that to give us this template here. Now, in terms of setting up the MCPS, so what you're going to be able to do within this is if I go over to setup MCPS, and again, if I just scroll down to the bottom here, once you're within here for the first time, they do actually have the MCP configuration all set up within here. The great thing with how it's set up is you can go and you can click to draft what the MCP looks like for exactly what I actually showed you. So say for instance you want to leverage Zapier like I just showed you. All that we need to do is we just need to copy this URL from the connection tab within Zapier. Once you've selected all of the different integrations, all that you need to do to leverage all of them is just paste in that URL right within here. Save it out. And then from there you can go ahead and click let's go and you'll be off to the races. As soon as you've added them in, what you can do is you can go ahead and click let's go and it will begin to initialize everything where you can leverage all of that. Now, the other thing that I do want to mention is within here there are a number of different shortcuts. But what you can do with all of these shortcuts is let's say you have conflicting applications that leverage these same shortcuts or if you just have preferences on how you want to actually invoke the command, you can go ahead and change out all of this. One quick thing that I did want to show you is they do have a really great quick start guide and within here there's a number of different examples. Say if you want to get CSS from Figma designs, you can go through this example and you can see exactly how to set all of that up. Now, the one thing to note with some MCP servers is they do have certain dependencies. Just make sure that you do have Node.js installed, mpm, MPX, all of that good stuff, and you should be able to get set up. In my opinion, where this starts to get really exciting is in the advanced flows. I demonstrated one of them where it took that document from my desktop, and it was able to break that out into linear, but you could imagine maybe breaking it out into linear and then sending an email, updating a Slack channel or what have you. You can play around with this for whatever use case you have in mind, but otherwise kudos to the team over at Deepgram. This is a killer launch. I encourage you to check this out. Try it out. There are a ton of really novel use cases for how you can leverage voice to both accomplish things easier as well as faster. I think this is just a glimpse into some of the capabilities of the future. And with it being powered by Deepgram Graham, we know for sure that it's both going to be fast as well as accurate in terms of the responses that we get. So otherwise that's pretty much it for this video. If you found this video useful, please comment, share, and subscribe.