Tof

Building tools I wish I had.

💸 : cut LLM bills 30%
🚇 : localhost to the world
🎨 : visual data workspace

[Tof @toflou]

Launching Tokenwise today.

I've got a few side projects running at the same time, and at some point their LLM bills turned into a number I just stopped questioning. Every month I'd add credits on OpenAI, then Anthropic, never really knowing which project or which change was eating the most. New models kept dropping too, so I'd switch things around half-guessing whether it actually saved me anything. Mostly I just paid the invoice and told myself I'd dig into it later.

What actually got to me was the spend I couldn't even see. Claude Code runs all day and none of it usage shows up anywhere until the invoice lands, and it's a real chunk of money sitting completely in the dark. Even with a Max or Ultra subscription, I thought it would be possible to optimize the plans to get more out of those.

This is basically the proof of concept I built into Tokenwise.

You add one line of code to get a proxy of all your LLM requests. After that, every call your app makes is just there. What it cost, how slow it was, how many tokens it ate, whether it failed. You can break it down by model or by whatever you tag.

The part I actually care about is what happens next. Several time a day it goes through your real traffic and points at things worth changing. Like, this prompt runs fine on a cheaper model, here's what you'd save, and here's the quality score so you know the output won't get worse. You click apply or you ignore it.

It sits on Cloudflare so the latency it adds is basically nothing. Your users won't know it's running ( It was one of my main concerns for the Datadef users !)

Most people I've shown it to are paying way more than they need to and never had a way to find out. That's the whole reason it exists.

It's live now, come test it

If you build with LLMs, try it and tell me where it falls short. Right now that's more useful to me than anything.

[Tof @toflou]

American people in Vegas are very ugly

[Tof @toflou]

@rownation47 @shipordie_ Don't be an attention whore, it's the game, it's rough, you have to be humble and keep going
You don't have to tweet to make it work

[Tof @toflou]

@Polymarket People just use https://t.co/u6Hp3ss1O4

[Tof @toflou]

@iFeyz2 Hello
C'est vraiment lourd et ça pourrait m'intéresser pour des futurs projets immo

Attention au scraping de tes sources, Jinka s'est fait condamné il me semble par rapport au scrapping de SeLoger et Leboncoin

[Tof @toflou]

I swear to god Claude Opus 4.8 is so much dumber right now

[Tof @toflou]

@TokenwiseAI is making it into the top 10 !

Please help me reach Top 3 Launch of the day with a simple upvote on @ProductHunt

[Tof @toflou]

@saen_dev Thank you very much sir ! Please feel free to give it a try for any of your project
We value effiency and privacy

[Tof @toflou]

🎉 Tokenwise is trending on Product hunt !🎉

Tokenwise: A smart LLM proxy that shows where you're overpaying

Link in the first comment !

Give me some strength and your opinion on product hunt launch page !

[Tof @toflou]

@marclou can you allow trustmrr to connect also with @gitlab for us the orcs that do not develop on Github

[Tof @toflou]

Been busy shipping, I proved myself I can make some $$ on internet with my first SaaS last winter.

I'm now going all in with Tokenwise, scaling it to 10k MRR 📈

[Tof @toflou]

We need a spending version of it, so here is my shot !

I spent $326,79 in May 2026 !

🧑‍💻Claude Code Max x20 - 180$
☁️Cloudflare Domain Registration - 40$
📩Resend API - 0$ (still in free limits)
💻OVH VPS server hosting - 6,79 $
🤖Open AI credits - 100$

[Marc Lou @marclou]

I made $87,507 in May 2026.

⭐️ TrustMRR — $27K
📈 DataFast — $20K
🏴‍☠️ Ship or Die — $20K
🧑‍💻 CodeFast — $8K
⚡️ ShipFast — $7K
🐥 Twitter — $3.2K
🦐 SuperShrimp — $1K
🍜 Indie Page — $715
🛡️ ByeDispute — $248
🎞️ YouTube — $111
🌱 HabitsGarden — $97
🧬 BioAge — $49
📚 WorkbookPDF — $19
💩 PoopUp — $18

To whoever bought PoopUp: Thank you ❤️

[Tof @toflou]

@MicroLaunchHQ my DMs are open !

[Tof @toflou]

Tokenwise is so cool, I can just see live the cost of my different SaaS, and get a breakdown per page/domain of my product

For example, I use AI in multiple different locations in my other SaaS, I can just tag each requests made by my app, and deep dive the requests, costs, latency, and trigger individual optimization for each of them.

So far I save 20$ per day in LLM calls for datadef diagrams generation, with only few rules activated, without any of my users noticing any quality change or issues

[Tof @toflou]

https://t.co/TEdctAvg1S

PewDiePie new Odysseus project sounds very promising.

Finally a good product to self host for AI usage.

Now, I need to buy 10x GPUs to have proper models running on my side.. 🙃

[Tof @toflou]

What is happening, some kind of AI Crawler/LLM bots that spam impressions out of nowhere with weird google searches

[Tof @toflou]

Claude Opus 4.8 cost and token spike detected within my Claude Code cli instance.

[Tof @toflou]

@JustJerry121 Tokenwise new update even allow you to proxy your claude code cli requests to tokenwise and see usage and optimization based on your usage

[Tof @toflou]

I'm looking for SaaS founders that use a lot LLM providers / AI call and that wish to monitor and track their LLM consumption

let's connect and work together

[Tof @toflou]

@charlieholtz No