Building a zero-setup API debugger, would love feedback
Hey PH
I'm working on toran, a live API inspection tool that works with just a URL swap. No SDK, no proxy config, no cert setup.
The problem - I couldn't see what my code was actually sending to third-party APIs. Debugging meant console.logs everywhere or messing with Charles/Proxyman certs.
toran is simple. You swap your base URL (like api.openai.com becomes ns262aajok1jv.toran.sh) and watch requests show up in your browser as they happen.
I've mostly been using it for MCP servers and LLM API calls but it works for anything.
Curious to hear from folks who:
Build AI agents or MCP tools
Work with third-party APIs (Stripe, Twilio, etc)
Have strong opinions about API debugging
What's missing? What would actually make this useful for you?
toran.sh if you want to try it (no sign-up)
Replies
I've burned way too much time fighting proxy certs, so this immediately caught my attention. Watching requests live in the browser feels refreshing. For me, response time graphs or basic performance metrics would make this insanely useful during tuning.
@kathrine_rolce Ha, yeah the cert dance is exactly why I built this. Glad that resonated.
There's already a timing breakdown that shows TTFB, transfer time, and toran overhead separately for each request. Not graphs yet but you can see the numbers.
Curious what you'd want from graphs specifically. Like spotting slow requests over time? Or comparing latency across different endpoints? Trying to figure out if this is a "nice to glance at" thing or something you'd actively use for perf tuning.
I work a lot with Stripe and random third-party APIs and visibility is always messy. I like how frictionless this is. Personally, I'd want redaction controls for senstitive fields before sharing logs with teammates or clients.
@magdalena_anderson Thanks! Yes, the goal was to make it as low-friction as possible.
toran already redacts sensitive values by default, and you can share torans with teammates or export requests as HAR files. So the sharing with clients use case is covered.
Would be curious to hear if that works for your Stripe debugging or if there's something specific you'd want different.