S2. Ep16 - Meet Clyde: The AI Business Partner You Can Actually Trust
What if you could have an AI business partner that actually keeps your data safe?
In this episode, Katie and Noel properly introduce Project Clyde an AI agent platform Noel built from scratch after losing trust in the alternatives. They walk through how Clyde builds and manages teams of sub-agents through simple conversation, why it sandboxes file access and encrypts everything via Supabase, and how the security-first approach compares to Open Claw's plain text storage.
They also cover Anthropic's Opus 4.7 release and its trade-offs versus 4.6, the new Claude Designs tool (and how Noel burned 67% of his weekly allowance in 20 minutes), the Mythos model gatekeeping, and the podcast hitting 10,000 downloads.
Want to learn more about Clyde, head to projectclyde.app, if you would like to join the free community you can find this on Skool (Project Clyde).
How to find us:
Join our membership over on Skool, where we support you on your AI and automation journey. We share exclusive content in the membership that shows you the automations we talks about in action how to build them. Find out more about the AI Business Club here.
We have a free LinkedIn group (AI Automations For Business), the group is open to all.
New for 2026, you can also find us on Substack, click here to subscribe and get all the latest news and updates from us.
If you would like dedicated help with your automations or would like us to build them for you then you can find our agency at makeautomations.ai
-
Katie (00:26)
Hello, welcome back to another episode. Hi, hello, I'm Katie, and as always, I've got Noel here with me. Hi, Noel, how you doing?
Noel (00:33)
Yeah, I'm doing great. How are you doing this week?
Katie (00:35)
Yeah, all good, thank you. So apparently there are a lot of updates within the AI world, thanks to Anthropic. So before we get into what we're talking about today, do you just want to go through some of the major updates and then let people know how it's going to affect them or their businesses?
Noel (00:46)
Yeah, so I guess the big update last week was they released the Opus 4.7 model. So that's something we discussed on last week's podcast. It was a leak as always. But yeah, within a week it was confirmed and released. And it's an interesting model. So again, it's supposed to be the most powerful thing ever, that flagship model, but they've kind of doubled things down, which I thought was quite interesting.
In certain areas, they've made it worse than Opus 4.6. And I think that's potentially so they released another model as well, didn't they, the other week. Mythos, which they said no one's going to be allowed to touch apart from big corporates, big software companies. The things that they dialled down were to do with security, compliance and things on the Opus 4.7 model. So I think they specifically trained it so it's a bit better, but actually far worse than the previous model when it comes to security compliance. So they make that Mythos thing the best. Really interesting.
Katie (02:17)
Do you think that's to encourage more corporate users rather than free or low tier paid? They want people having that corporate package almost.
Noel (02:26)
Yeah, I think so. But with Mythos though, they've really gatekept it to a certain number of companies. So I don't think it's something you could apply for and then get access to. It's kind of like, these are the big players, these are the only ones that are allowed to touch it. I don't know if they might update that in the future for enterprise users and things like that. Maybe.
Katie (02:52)
Interesting, very interesting.
Noel (03:03)
Also on Opus 4.7, myself and a few others in the AI space have been having a chat about it like we normally do. We've noticed it's also maybe not as good generally as Opus 4.6 when we're having chats within conversations. It's hard to put your finger on it. It doesn't feel the same. It's a bit weird. Not to say it's rubbish, it's very good still, it just feels a bit different.
Katie (03:39)
But none of you can say why it feels different.
Noel (03:43)
Some people have said online that it's not as clever and things like that, but I don't think it's that. It's kind of like the tone and the response has slightly changed. I'm sure I'll put my finger on it soon. I use it every day.
Katie (04:02)
Interesting. Any other updates?
Noel (04:05)
So the other big Anthropic update this week, and they were really the only ones that did anything in the last seven days. They released a new tool that you can use as a design tool. So if you're designing maybe presentations or you've maybe coded an app or a website, you would normally use something like Claude Code, which doesn't give you that visual representation and brand designs and things. You kind of wait for it to be done, have a look, see if you like it, and then iterate from that point.
But with Claude Designs, it's different. It's a completely separate platform from everything else. You can't access it via the chat. I think it's claude.ai/designs to get to it. Essentially, you can upload your website. So if you've already got one, it gets all of your brand guidelines, all your colours, all your fonts, all the sizes, all that sort of stuff. It'll gather all the information. And then from there, you can design presentations, apps, or even redo your website while keeping the same brand styling.
There's another app widely known with UI developers called Figma, which does similar things. But most people online say this completely kills Figma, which it doesn't. They're two completely separate things at the moment.
But they have separate usage as well with the designs versus your normal usage with the chat. And what I found is I had a chat with it for about 20 minutes. I'd used 67% of my weekly allowance in designs. So it's not like right now you could deep dive into a project and go absolutely crazy and keep iterating and changing things because you're just going to burn through all of that usage within an hour. I mean, this is the first iteration of something. So I'm sure it will get better.
But I think Anthropic get quite a lot of grief at the minute for their usage policies, how many tokens, and I think they've reduced it recently as well. They've reduced the caps slightly. So there's quite a few unhappy people about that.
Katie (06:40)
Okay. And is that it on the updates? Surely there can't be any more.
Noel (07:01)
They completely redid the user interface for the desktop app. So over the last three weeks, every single day I've opened it, there's been an update. Sometimes it's multiple times a day. But what I was starting to notice is they started changing colours a little bit. And I was like, that's frustrating, I don't really like those colours. I like the ones that were there before. And then the next time I clicked update, it completely changed to a completely different theme.
Claude Code's changed as well. So you can now view code files if you really want to. But we don't want to get too nerdy.
Katie (07:34)
Just the design you mean. And do you think that's leading to something new?
Noel (08:00)
I guess so. It's moving more into being able to read and then edit the code yourself within the app. So you're not having to use two different apps. You can just use the Claude app to do it. There's definitely that shift. Whereas with Cowork and chat, they're widely the same really as what they used to be. They just look different.
Katie (08:39)
Okay, good to know. That was a lot of updates.
Noel (08:43)
Yeah, I'm sure there'll be even more next week.
Katie (08:46)
Probably. Now, as always, we love to keep you updated. If there's anything that you think we haven't talked about and it was an update, then please let us know. We'll definitely talk about it. Let us know if there's any other updates that you've been aware of and you think our listeners should hear about it too.
So in this week's episode, we are talking about Clyde. And if you've never heard of Clyde, then don't worry because this is something Noel has created himself and he's even named it. So Noel, let's start at the very beginning. What is Clyde?
Noel (09:23)
So Clyde is an agent platform, which you can basically chat away with. And essentially what he does is you can build a team of experts to work with you on your company tasks. We can do all kinds of different things, connect all kinds of different apps. And essentially it's kind of like a one-stop shop for you and your business. We can connect to it anywhere in the world using things like Telegram. Soon you'll be able to access it via a web app.
I basically see Clyde as a business partner for you and your business, especially if you're a solopreneur like myself. He sits there and he does all the boring stuff that I don't really want to do or haven't got time to do. That's kind of the premise of it.
Katie (10:15)
Okay, so it is mainly aimed at solopreneurs, people who are running their business on their own. And what about for people who do have a small team, is it going to be suitable for them as well?
Noel (10:20)
Yes, it's suitable for any company size. Right now, there isn't a way of having a team account where you could all interact. We're not at that point yet. But that's something I would definitely be looking into. It can be for anybody. You don't have to be a tech expert to use it. It's just a simple chat interface. You don't have to be super nerdy in order to get it to work.
Katie (11:02)
Okay, so can you tell us what are some of the things that it can actually be used for and why you might want to use Clyde?
Noel (11:14)
So Clyde is essentially your business partner. Let's say you want to have a content creation team. Normally you would have to go off and hire those in the olden days, whereas now Clyde can basically hire and manage those agents for you. They'll go off, they'll do tasks. We can schedule them to do tasks and essentially it just makes it so much easier to manage all of these agents from one single interface.
There's plenty of other things out there that do similar things. There's apps like Paperclip. There's also Open Claw, which came out earlier in the year, which does some fairly similar things. But with Clyde, it's been built so it's more secure for you and your business. You can just chat away and build all these teams.
Normally you would have to create all the prompts and things for your sub-agents. Don't have to do that. Clyde does all that for you. You just say, I want this team. This is what they need to do. These are the goals that I want them to achieve. And it will go off and do everything else for you, which is pretty awesome to see.
Katie (12:27)
Yeah, so what sort of things can Clyde actually do then? We know it's in charge of all the agents that you're using within your business. You mentioned that it can be a content team, but what about if someone is like, well, I'm in charge of my content and I'm the face of my content. I don't want anyone to touch that. What other things can it do for you?
Noel (12:56)
So it can connect to any application currently. There are API and MCP integrations built in. So you can connect to things like your databases, your Google Sheets, your inbox with Gmail and things like that. All of that is pre-configured and ready to go straight out the box.
You have to think of it as, if I want an agent in my business today, what would I want it to do? And basically, Clyde could do that for you. We're not having to create and connect workflows within tools like Make.com or n8n. We're not having to code anything. Even if we're unsure how to connect to a certain application, we can just copy paste the link to the documentation and then say, right, Clyde, sort this out, please. And it will go off and do it for you. It will create all those integration connections. And then it will just go off and do those jobs for you.
Katie (14:00)
Within Clyde, how easy is it to create an agent? Because obviously we've talked about how to create agents using different software platforms and there is some tech involved. A lot of it is communication and figuring out what you want it to do. But with Clyde, how much of it is tech and how much of it is actually just communicating?
Noel (14:05)
I would say the way I've built it and the way I've always wanted it to be is 90 to 95% communication. I don't want it to be a barrier where you're stuck on the technical side of things. It's just like you would have a chat with it, like me and you having a chat now. I say to Katie, look, I want to connect to this. Can you go and do that for me? And you go, yep, sure. I can go off and do that for you. That's as technical as it needs to be.
I guess the last 5% is getting API keys and credentials and things like that. I wouldn't class it as technical skills, but you would need to go off and know where to find those things. It's definitely conversation first.
But we have built in some other bits as well which make it maybe a little bit more accessible for the engineers. If they want to put in certain API requests, they can see it and see a bit more history of what's going on. So there is that element as well, but for most people, you just talk to it. And it can also assign all the integrations to sub-agents as well, which is pretty awesome. So every time you run that sub-agent, it knows all the tools and how to use them. And it will go off and do those jobs for you.
Katie (15:55)
So what would be the advantage of actually using Clyde for your business?
Noel (16:01)
So I guess the advantage is having that central repository to sit down and chat with this agent. It's completely configurable to you and your business. There are things like how we have in Claude, we've got skills. So we can create our own brand voice skills. We can create skills for specific processes that you might have within your business. So it's completely customisable. It's just up to you as the user to figure out what you would like to use it for.
Katie (16:35)
And how easy is it to manage when you've got all these agents and sub-agents up and running?
Noel (16:44)
So it depends how big the team is, but really you don't have to manage the sub-agents. You don't have to worry about those in day-to-day running. They just get called upon by Clyde as and when a task comes in or whatever you've asked for.
But what I have done is collated them into teams. So you could have six agents which are part of your finance team. They would have a little finance block within the organisation chart. And then you can click on that and see the structure of the organisation underneath.
Before I added that in, all of the agents were on a standard corporate organisation chart, which was just getting out of hand and too busy. So they are condensed into teams. It's easy for you to figure out and see who's doing what. You can have a look at what skills they've got access to. You can edit their system prompts if you want to. All of that is open and available to the user to edit if there's a need to.
Katie (17:53)
And we spoke about, it's had so many names, Claudebot and Open Claw. And what was the other name?
Noel (18:02)
Open Claw, Maltbot. Maltbot was the middle one, yep.
Katie (18:22)
So obviously you had reservations around using that particular software because of the security issues. What makes Clyde different to Open Claw?
Noel (18:28)
So the difference is, especially with the... I've got two versions of it. There's an open source version which you can download and host on your laptop, computer, whatever. And there's also going to be a cloud version.
With the open source version, the agent is constrained to work within the bounds of the application structure. So it can't go off and save and get files from your computer. I've created a working area which is kind of like a file system. You can share files with Clyde and Clyde can save them in there for you. You get to see that in the user interface. It can't get out.
For the cloud version, everything's encrypted. All of your messages are secure. Whereas with Open Claw, I don't know if they've changed it yet, all of your messages used to be stored in plain text. So if somebody knew how to get into your computer or virtual server and knew you had that installed, they would be able to find that information. And if you've been sharing any personal information with your agent, they could just read it. That's not the case with Clyde.
With the open source version, we've also linked it into Supabase. So that's a database, but nothing of your chat history is stored locally. That's one of the big things I wanted to make sure of. I don't want to be liable for anything like that.
Katie (20:23)
Yes, because you were inspired to create Clyde because of all the issues that were going on with Open Claw around the security issues and people were putting it on their laptops and having all the data and client data looked into and taken.
Noel (20:42)
Yeah, although we're giving it access to the inbox. There was a lady from Meta, wasn't there? She was part of AI and compliance and she gave access to her inbox and it started deleting everything because it thought, well, that was yesterday's email, so I'll delete everything after yesterday. Yeah, it can go a bit rogue.
But I think for me when thinking about Clyde, I was looking at Open Claw and it was like, I like the idea of it. I like the premise. I just couldn't trust it enough to have it with my data or put anything client based through it. So as a business owner, it was a big no-no.
When I built Clyde, I was like, right, I want this to be something that I can use every day. And I want other people to be confident enough to use it every day for their business work. That's kind of where I was aiming for.
Katie (21:47)
And do you feel like you've done what you set out to do?
Noel (21:53)
I believe so. I'm pretty happy with how it works, how it's all constrained. In the early days, we talked about this weeks ago on the podcast when we first started creating it. I kind of wanted it for the open source version to just work within the app directory and bounds. And essentially, I forgot to give it permissions to write to the working folder and create folders. So it was just like, I don't know, problem. I'll just run this command and then save it somewhere on your computer.
And I was like, this is exactly what I was going to narrow down. But it just goes to show how clever these agents can be. It hit a problem. It was like, hold on, I could just figure this out and do this instead, which is very clever. But it didn't break out after that point, which is good.
Katie (22:50)
So how do people actually get started with Clyde? If people are thinking, this sounds great, I want to start using it within my business. How do people get started?
Noel (23:05)
So the website is projectclyde.app. On there, there's links to get the open source version. You'll be able to download that from GitHub. They'll take you straight to the GitHub repository. All the setup instructions are there to help you get started. If you do get stuck, just give the readme to an agent and it will be able to help you out and set things up.
Katie (23:35)
And you've also got a community for it over on the Skool platform, haven't you?
Noel (23:35)
I do, yes. When I released the cloud version, I wanted the community to be a community of users. So it's a mix of the open source and cloud customers. You can all come together. It's free to access, we'll leave the links below.
There are a couple of paid tiers in there as well. So if you want to support the development, the first one's like $5 and basically you get extra early access to new developments on the open source. You also get some weighted privileges when it comes to voting for features. So if you're on the premium tier, you get a times two. And if you're on VIP, you get a times three. So you get three votes, technically.
Katie (24:35)
It's starting to sound like a game of Traitors. You get two votes.
Noel (24:44)
Exactly. I wanted to build that so we could all come together. I don't just want to build this and then have people using it and not have a way of talking to me and talking to other people about it. It's a way for us to all come together and say, well, it'd be cool if it did this, and then have that chat about it. As well as extra support in there for people.
Katie (25:16)
Yeah, I think it's always nice to find out how other people are using it or what they're using it for. It's kind of like when we always say to people, if you're using AI or automations for something that isn't just run-of-the-mill stuff, then please tell us because we'd love to hear all about it. We might even ask you to come on the podcast and tell everyone about it.
But yeah, it's always really interesting how people are using different pieces of software and AI.
Noel (25:49)
Yeah, it's crazy isn't it? Some people's imagination on how to use AI is brilliant. We use it every single day and there's still things now that I see and I'm like, I don't even know why I didn't think of doing that. That's such a clever idea.
Katie (25:57)
Yeah, how other people's brains work is amazing. And if you think we're talking about you, then please get in touch with us. hello@makeautomations.ai. We'd love to talk to you about what you are doing, what you're creating. And if you're up for coming on the podcast, definitely let us know. We're always looking for people to come on and tell us what they're up to. And I think it's really encouraging and inspiring to find out how people are using AI or automations.
Noel (26:42)
Yeah, definitely.
Katie (26:43)
And Noel, we've also got a bit of news, a bit of congratulations because our podcast has now hit over 10,000 downloads. So thank you to everyone for listening to this podcast. Thank you to you who is listening right now for downloading, for listening. If you haven't left us a review yet, please do. We really appreciate it and it helps us to grow this podcast.
Thank you so much, Noel, for telling us all about Clyde this week and for all of those updates. And we will catch you for another episode very soon. Take care.
