gsundeep

Agreed, I think adding guardrails to this would be really useful to ensure the AI only has limited permissions to these services (or asking for some sort of confirmation before making potentially dangerous tool calls).

Currently we are only recording which tools were requested by the MCP client. We don't store details of the executed tool, neither the arguments nor the response. Currently we are not open source but we are considering that. Thanks for the feedback!

We are thinking of open sourcing it, the current codebase requires Cloudflare Workers so it will take some changes to make it more generic. Thank you for the feedback!

We’ll prioritize adding support for this client - thanks!

We'll soon be adding the ability to have multiple models perform the scan in parallel, so any attack would have to bypass all of the models.

Thanks for your comment - MCP Defender sits between the MCP client and server, it doesn't need to worry about the protocols that the server communicates with to other services.

This is certainly a valid concern. We'll soon be adding the ability to have multiple models perform the scan in parallel, so any attack would have to bypass all of the models.

With the default signatures, source code would not be treated as malicious. However, you can add custom signatures and detect whatever you'd like. We'll soon be adding deterministic rules as well to complement the LLM based ones.

MCP Defender sits between the MCP client and server. If you use Cursor for example, MCP Defender rewrites your Cursor MCP config file so that all MCP servers point to the MCP Defender proxy. So the tool calls are scanned before they make it to the server. The responses from the servers are also scanned although this is configurable (disabling it speeds up scans).

This is really interesting, I'll check it out. At least in its current form this seems like it would take some effort to setup - we're focusing heavily on making MCP Defender easy to setup in less than a minute and then forgetting about it as it runs in the background.

We used Cursor + MCP tools like Cloudflare, Linear and Github to build and deploy a lot of MCP Defender, so I think the value is real. I had the same thought about it feeling like an antivirus/firewall many of us ran decades ago. Those always felt clunky and slowed down your computer. We'll try our best to avoid that fate

I had the same thought while building this, but I really feel a tool like this is needed as MCP has a lot of surface area for attacks. Any MCP server that gets hacked exposes all users of that MCP server to serious security risk, unless they are really careful about inspecting every single MCP tool call they make.

While Cursor and other apps can include security checks in their system prompt, MCP Defender provides an extra unified layer of security across all apps. Also, we're going to be adding the ability to have multiple models perform the scan in parallel so any prompt injection attack would have to work against all of the models you select.

We'll be adding the ability to run MCP Defender through a local LLM soon, so using that approach no data will leave your computer to perform a scan.

Yes, there is a delay for MCP exchange, but I imagine that most MCP calls in the future will be done in "YOLO" mode where the user prompts a large task and an agent makes 1000's of MCP calls over hours to accomplish it. This would add some time to the overall task but IMO this is a small price to pay for added security. Also, the delay will decrease over time.

We're planning to add deterministic rules on top of the current LLM based ones

Thanks! Yes - which client are you using? We'll add support for it

Not yet, but that sounds like a great idea as watch apps are often a lot simpler than iOS ones. I’ll put that on the roadmap - thanks!

Sorry about that - can you try one more time? I just added some additional logging

Spawn apps can use the accelerometer - thanks for pointing this out, I'll clarify the website.

I pasted your prompt and got this: https://imgur.com/a/3Wg8AuK

During the recording, I moved around around and stopped twice. The only thing not working is the audio volume changing based on the g-force readings. I'm digging into why that's not working.

I totally agree! When you export your app in Spawn, it’s exported as an Xcode project and you can make code modifications from there

I made some slight modifications to your prompt and got something working:

Here's the prompt: https://i.imgur.com/lUjLTsd.png Plain text: https://pastebin.com/HTUjcy7M

The two differences from your prompt are I changed "A listing of horses" to "A listing of horses that the user has added" and at the end I added "Use NSUserDefault for data persistence, not coredata". I'm modifying the model prompt as something like that should definitely not be needed in the app specifications.

Some screenshots of the generated app: https://i.imgur.com/IwEKuZW.png https://i.imgur.com/zONc9Fm.png

Thanks for giving it a try! I just took a look at your prompt and it looks well defined. I'll investigate the generated code and reply back here.

That's a fair point, I can't say for certain that tools like Spawn will be able to build really complex software. However, I think there's a ton of software that could be built with current model capabilities that simply isn't getting built because the remainder of the process still has a high learning curve (ex: learning how to use Xcode, figuring out app permissions etc.)

This looks awesome, any thoughts on making this automated?

Here’s a similar product I ran into here a while ago that uses the info in the other meta tags (title, description) to automate the creation of these images: https://www.mugshotbot.com/

I also built something similar called Thumblink, which takes a screenshot of the page to create these images: https://thumblink.com

Here are some examples:

If Stack Overflow used Thumblink: https://imgur.com/a/RVMRLxD

If Substack used Thumblink: https://imgur.com/a/eOn4C0y

Hi HN,

I started working on Thumblink while trying to generate link thumbnails for a past project.

Most websites don't configure link thumbnails, which usually results in ugly links. Thumblink handles the creation of thumbnails and configuration in one HTML tag. The generated thumbnail is a screenshot of the page, giving potential visitors a taste of your website before they visit.

Check out the demo: https://thumblink.com/demo

Would love any feedback!

Do you have any info about what it's called or how it's used? I'm surprised ESPN doesn't have a clickable video link for their play by play for example. NBA has it for league pass holders, but they also don't offer it in real time, which to me defeats the purpose. I want to see and share a video of something that just happened, not only forced by having to wait until after someone else covered it in a highlight.

Hey HN - we built Clips.live since we wanted to share clips of NBA plays as soon as they happen. Before this, you had to make these clips yourself or wait until someone manually made them in order to share them. Eventually, we want to record and index every play of every live sports game.

Only the user typing “assist” needs to have a subscription and they can use it across all of their machines on a single license.

I’ll clarify this on the site - thanks!

Sorry for the lack of clarity - only the user typing “assist” needs to have a subscription. We host publicly accessible tunnels so that this works wherever you are.

I’ll clarify the subscription details on the site - thanks!