Show HN: I built a smart proxy so your coding agent can run loose

ddandriscoll 1 day ago 6 commentsRead Article on trollbridge.dev

ZH version is available. Content is displayed in original English for accuracy.

The only way to go fast is full YOLO mode in your coding agent. I've got the local sandbox figured out (pro tip: Incus VMs work great) but I wanted to keep my agents from doing things like inadvertently blowing up my cloud services or chasing a prompt to POST to some random website. I struggle most with this on my side projects where my permission model isn't quite as robust as it is at the office.

I started with a firewall on the Incus container but every time the agent needed access to something new, I was poking more holes in it - and it didn't differentiate between HTTP verbs.

I've been using Claude Code auto mode too but I wanted to be able to guide the LLM, not just hand over control completely. I also want some flat allow and deny rules.

So, I built Trollbridge: an HTTP/S proxy that runs in front of your coding agent and lets you set up allow or deny lists, realtime approve or deny in the TUI, or wire up an LLM that makes calls based on your prior decisions.

Would love to hear what you think - give it a shot, leave a comment, file a bug. Thanks!

Website: https://trollbridge.dev Github: https://github.com/dandriscoll/trollbridge

⚡ Community Insights

Discussion Sentiment

100% Positive

Analyzed from 103 words in the discussion.

Discussion (6 Comments)Read Original on HackerNews

nzjrs•about 3 hours ago

Your entire website was written by Claude and reads like slop. If you can't write a readme by hand then I don't trust your code.

tln•about 1 hour ago

Yes, but it's a nice website. And it looks like a really nice tool.

Given the effort you've put into the comment I doubt you looked much at the code, but the github also looks well-crafted.

FWIW the README is actually way better curated than the average Claude-only README

transreal•about 2 hours ago

This looks very similar to Docker Sandboxes.

philipwhiuk•about 3 hours ago

Hopefully your advisor LLM doesn't make calls of it's own.

dandriscoll•about 3 hours ago

LLM only, no tools. It's got a brain but no hands ;-)

grim_io•about 2 hours ago

Can the advisor prompt inject the agent and steer it in the direction of world domination, or alternatively gaslight the agent into rm -rf /