Claude & Codex control your computer

The Next New Thing· 16 min· 3,094 words· 14 min read· English ·Watch on YouTube

This is the full transcript of Claude & Codex control your computer, published on YouTube by The Next New Thing. Every paragraph carries the moment it was spoken, so you can click any line to jump straight to that point in the video, search the whole thing for a word, or copy it out.

0:00Vlog code is now controlling my browser through the terminal. And if you use the desktop bridge, it can also control my computer and I'll send you a text message right now. This will be available to anyone who's watching this. They'll be able to use this. Yeah, we'll have a link in the description. I'll do a screen share so you can see how easy

0:15this is and I'll just show you. Presented by Zapier, the AI automation company. Let's see how this works. Basically, you just log in here. I already have the MCP connections established both the Zapier and the browser bridge and the desktop bridge. So I can say something like use the Zapier MCP to look at the conference

0:39sheet. I have a list of conferences in the demo tab, which I made specifically for this demo to make it easy. Um please help me apply to speak at them using the browser bridge. And this in browser bridge is what you created to give Claude code access to your computer. Correct. Within the terminal. Yeah, and technically even Codex really, but any any AI that has the ability to connect to MCP servers

1:12in any instance anywhere really. Anywhere they can they can connect to a local MCP server. Okay, and by the way if it can connect to the browser, why are you bothering with the Zapier MCP? Why don't you just tell it click my browser, go find this file and let's go. Uh it's much faster this way. So anything that I can do through MCP servers or API endpoints, I still will. But in terms of like filling out forms, getting information that is unavailable through APIs or um actual integrations, this is great for that. And also doing actual work like this is a good example. Reaching out to

1:51people on LinkedIn. Um and and also if for vibe coders, full front end actual testing is one of the best use case for this. So I will use this for like actual genuine work like I am now. But it also makes my tools that I vibe code just a lot better because it can do front end testing. It can actually see itself. So if I wanted to fit a brand design as an example, it doesn't have to guess by just looking at the code or a single screenshot through something like play right, but it can actually click around and do full front testing. So it makes

2:26everything I'm doing on the computer easier and better. Okay, I can see totally how this would be helpful. Um all right, let's see what it filled out. All right.

2:49Whoa, it's controlling the browser. It's actually filling it in. I could totally see myself using this. Yeah. I'm so excited about this. This is so weird that I get excited about this that it's like hype. And it's just my computer doing stuff. And it's just doing it. It's more superpower for me and my agents. All

3:05right. Let's let's come back and look in the terminal and see what happened while all the browser stuff was happening. So this is what it looks like when it's done. And you can see for purposes of the demo, I had it replace my actual phone number with 555. But it has my real number and everything. It it the way that I do this is I have one folder for speaker applications, I have another folder for LinkedIn outreach. And then those folders have the relevant information that it needs. So the speaker application folder has my bio and all the history of all my prior speaking

3:39engagements that I can reference. Okay, and it's basically just saying I went through it, I filled in the forms, everything is good and now you decided that the way you want to handle it is have the forms filled out, but only you can go and submit them because you want to make sure that everything's filled out properly. Correct. Yeah, so often times it'll say that it needs maybe one or two attachments for me to fill out. It'll give me sort of the last steps and then I review it and submit it. It's been

4:06working well for me. All right, so why don't we see how it can control your desktop? Sure, so I also this is the browser bridge. I also built also built a desktop bridge and that allows an AI through MCP server to control your actual desktop. So I message, opening new applications, resizing screens, things like that. How do you use it? What are some of the things that you use it for? Really, the main reason I built it honestly was for to see to for fun honestly. But also for I message because that's a big gap for me. Like I a lot of times I'll ask this

4:46to reference client emails or things like that, but or calls, but text messages are the gap. So now I can just tell it to reference text messages or send text messages to clients who prefer them. So that's the main reason. And then also with remote control, which Claude code now has, you can basically sort of use it as a your own open claw. You know, like you can use remote control now with these two apps Yeah. control your computer and your

5:13browser. And I trust Claude personally more than open claw. So that's one of the the other reasons. I could see, you know what? I was getting together with a few people here locally and they don't check email or if they do, they check it maybe a week later. Text always gets a response quickly. I could imagine myself saying, you know what? I'm going to schedule out text messages to all the people who I want to show up to this event. Maybe half hour before it'll go out on I message so it's blue bubble and I'll

5:42have Claude just send out the message. Even something like, hey, I'm heading over there now, looking forward to seeing you. Boom boom boom boom boom, it sends it out from my phone on my I message, it gets a phenomenal response. And if they don't have I message, obviously it goes in text. So let's see what that looks like. Can you share your

5:57screen that way? Yeah, absolutely. All right, let's see desktop bridge now. All right, cool. So You got I message on the left, still got terminal on the right. Yeah, this is the gravity terminal. I'm just going to say please use the desktop bridge to send Andrew Warner a text that says this is a live demo.

6:33Bam. I got it. Wait, no, did I get it or is this the last one? I got it right here. Nice. Perfect. Show it on my screen. Can you see it? I well, you're in my other space so I Okay, that's all right. Hopefully our viewer can see it. The interesting thing is I thought it was going to just go into that window and start typing, which is the way that Claude co-work works, which is sluggish and all kinds of

6:58clunky. How did it do it without typing anything in? How did it do it like that? So honestly, I'm not 100% sure, but I believe it uses the iOS SQL database that I that I message has. So it will just work directly with the I message database. So you're saying you built this and you still don't know exactly how it's doing

7:18it. Yeah, 2026. 2026. All right, we're going to have links to this below. Why don't why don't you stop sharing screen for a moment? Let's just talk big picture here. What are you building? What are you doing here? [snorts] Sure, so So you you give me more use cases to excite me for what we could do with this and then give it to me in perspective because you've been doing so many other projects. This is two of many that you have. Yeah, so I build a ton of AI tools for myself, for clients. I have a PR and SEO agency that I initially started building tools for

7:48and then it sort of has morphed into its own thing in terms of the AI practice. And I noticed that when I was using Claude, the desktop app of course has Claude in Chrome and it's good. It's it's it's sluggish and it definitely has limit a lot of limitations, but I noticed that my code was a lot better when it could check itself always, you know, including in Replit with Abacus deep agents. Always every tool that was able to check itself

8:18was better. And my biggest complaint about coding in the terminal was it didn't have that quality assurance check. And so I was thinking and actually I was watching a video of Boris Cherney. This is funny. Yeah, I was watching the the Claude code creator and he said that they used Claude code to create the Claude Chrome extension. And then that got me thinking I have access to Claude code. It's now six months later from the version of Claude code they used for, you know, their Claude Chrome extension. So maybe

8:50I can just do the same thing. And so I made my own version. AI called it browser bridge, which I went with. And then anytime that it it hit a road bump, I would go back and I'd fix it. So if for instance it was trying to work within Replit or within LinkedIn or within certain hard to navigate pages, I would go and I would say, hey, there's an issue. And I just went back and forth back and forth back and forth between the AI actually using the the MCP and then and the AI that was creating the MCP until both were happy basically. And

9:30now there's like 180 something tools. It also has in addition to just being able to navigate the browser, it has some SEO audit tools, some web security audit tools. It can look at the console. So it really is a tool built for coders and people who are trying to get work done on the computer. I can totally see it for annoying sites like LinkedIn where I just need to go and fire off messages, but also beyond the the day-to-day personal assistant

9:55stuff. When you create Claude code will create ugly things and not realize that what it did was not what it what I wanted. So, the move is you take a screenshot, you say this is what I see, what I wanted was something else, and then it goes, "Oh, now I get the problem." And it could adjust and correct. Now, what you're saying is part of your process of developing is to say, "And use browser bridge to check to make sure that it's looking right." And it

10:18could then go do it on its own. A huge huge part. And also, there's a lot of really annoying authentication issues with like Google as an example, or even Zapier is not annoying at all. It's very user-friendly, but I will still have my browser bridge use Zapier. So, I'll say, "You know, I logged into Zapier, set up what you need to on Zapier, set up what you need to with authentication or get the API keys you

10:41need. You have my permission." Um and then it will do and set all of that up. Um or if it needs to debug things, it will be able to debug things on, you know, the computer side, like the authentication API side as well. Uh so, it's really massive. Um and actually, a good example, um which is uh relevant because this is also about the browser bridge. I'll I'll share my screen here. Um I used the browser bridge to create this page about the browser bridge. Um and you'll see how in my opinion great it looks. Like uh it's it's our exact brand color, our font, and it looks good. And the only reason I

11:20was able to get this looking And this actually, by the way, is a bunch of other tools that we need if Anyone's interested. You need so many. Yeah, a lot of a lot of tools. is not even your business. You're in the business of helping other customer of helping your clients get attention, get PR, get traffic to their sites. This is like a a passion. Yes. And uh by the way, also a lot of Zapier agents. Yes, for sure. It's agreed. Um uh it's it's very fun for me. And the only reason it looks this good is cuz I used the browser bridge. Uh and it was

11:49able to continue editing until it matched the font and the brand colors perfectly. Uh and this also has more about the browser bridge. We actually have a few more tools in it now. I think it's up to like 186 or something. Uh although I should keep it for Blink-182 purposes for millennials. Um Uh but yeah, by the way, considering how much you're able to build, what is it

12:11that you use Zapier for? Uh a lot of integrations. Uh the MCP servers are are is still the main a main thing, and Zapier agents itself is just so easy to set up. Um so, you can see here Like what? What's the Zapier agent that you're using? Mhm. Uh a bizdev call prep agent as an example. This is one of the first things I ever set up, one of the first AI agents I ever set up, and it's been so useful. I I kept it. Um so, anytime a call is scheduled uh on my business development calendar or with the word with certain words in the

12:41calendar um that indicate it's a it's an bizdev intake, uh I get a it does research, and then I get via Slack a prep memo on who they are, why they might be interested in calling, um if it's for a if they're interested in SEO, it'll actually do a little SEO audit for them. If they're interested in PR, it'll tell me the last uh press releases they were in. Um I have an email bounce-back recovery agent, so anytime an email bounces back in as a PR firm, uh there's a lot of emails that bounce back uh with reporters. Uh so, this helps monitor all of those, and then if there isn't bounce

13:14back, it'll search for their new It'll both search for their new job if there is one, and also their LinkedIn page cuz their LinkedIn page doesn't change with their job. So, um it's a pretty reliable way uh to contact someone. Why are you doing it in Zapier? Uh it is just easier. Zapier agents is just very very easy to set up. Uh it has all these integrations, and like I mentioned, it was also the first on the scene. So, a lot of what I do now is with Zapier MCP, um but it's just so easy, and it was such a leader that I set all of these

13:47up, and now I use them every single day, so I don't uh plan on switching. What I love about it, by the way, is And obviously, they're sponsor, but I don't do any heavy-handed ads for them. Most times, people just see it in the bumper and move on. But I should say, the thing that I love about a lot of what they do is you could just talk to talk to them the way you would to Claude and say, "This is what I'm looking to have

14:07happen. Just solve this problem for me." And then yes, it shows you a flowchart if you want it. Yes, it shows you the process. But um but you're talking to a thing that's creating this whole workflow for you, and because it has connections to everything, it's got trouble-free integrations, and things just freaking work without having to come back and babysit them. I'll also say that they're just on the front line of everything. You know, like I mean, you and I, we talked about this. We were both You were apparently, I think, one of the first maybe Zapier customer. Um I was one of the first, and I think that

14:36probably uh is a good reason why we're so interested in this also. Like there's certain people and certain companies that are just focused on automation, and the rest of the world really caught up, but um I expect that they're going to continue being on the front edge of a lot of that. I freaking love all of this stuff. I absolutely do. Like really, I'm sometimes out in the backyard learning to play the guitar, and my wife goes, "Maybe the reason you're not learning to play the guitar fast enough is because you just keep watching YouTube videos about people who are showing you cool things that you could do your desk and

15:04And you know, last night she said it and I was so pissed. I go, "You do not understand this passion. I could actually do two things at once and all that." And then in the middle of the night I said, "Actually, maybe I should just put it away for a little bit." But it it's great because every time I learn something, like you just showed me now, I have another superpower, and then I think about, "Well, I do want to mail my book to a few different people, but I hate the USPS website. Now I don't have to care about it. I actually just use

15:27browser bridge." Mhm. You know, there's so many things that I want to that I want to do, and now you've given me the power to do it. Again, as always, everyone, we got links below to all of these things. Evan, thanks so much. For sure. If anybody has any questions or just wants to geek out about AI, I am very easy to get a hold

15:42of and would be very happy to do that. I love it. Thanks. Bye, everyone.

Where these words come from. This is the caption track YouTube holds for this video, written automatically by YouTube rather than by the creator. We read it, tidied the line breaks and laid it out so it can be read. The plain text version is at https://viewrankai.com/tools/youtube-transcript/b89XA9OPdvg.txt.

All rights in this video belong to The Next New Thing. Watch it on YouTube. If this is your video and you would rather this page did not exist, tell us and we will remove it.