LLMCap – A proxy that hard-stops LLM API calls when you hit a dollar cap | Hacker News Reader

LLMCap – A proxy that hard-stops LLM API calls when you hit a dollar cap

llmcap.io

26 pointscfaruk19 comments

Open articleView on HN

I've got a local one you can to do this for free with through openrouter

Even has a GUI.

$ uvx capit serve

https://github.com/day50-dev/limit-model-spending

* Free to use

* Everything is stored local.

* Anthropic skills are included for easily extending it to additional tools or providers.

Please ask for features or file bugs. Thanks!

Btw, I'm terrible at marketing even when I'm giving stuff away.

Proxying == Logging suspicions

guard-sdk does the same locally https://guard-sdk.js.org

Valid concern. Your provider API key is never stored — it passes through as a header on each request and is discarded immediately. LLMCap only stores token counts and costs per request. No key retention.

Its basic function is something you could vibe code in a few minutes :-)

Not to mention the whole privacy issue.

I actually laughed, its a security issue, and it requires payment to avoid payment.

Fair point on the irony. The 3-day trial is free — most people know within a day whether it prevents a bill that would cost more than $19.

I wonder much time did the author actually spent vibe coding this. I bet 90% of the time were spent on subscription payment processing.

I think the fundamental problem with this project is that if I'm on a hard budget, I have a hard time justifying $19–49/month to keep to my budget.

This will fail product-market fit instantly. Nobody is paying $49/mo for a tiny script they can vibe and run locally.

The pricing is very steep. I can see a market for this as an application with one time purchase but as a subscription?

We self-host LiteLLM which allows to set the budget per day/week, and it's free.

LiteLLM is a great option if you're comfortable running infrastructure. LLMCap is for teams who want enforcement without self-hosting or maintenance overhead.

Openrouter offers per API-key spending limits.

Came here to say this. And you get access to lots of models!

> The Service is provided "as is." We are not liable for downtime, data loss, or errors in token counting or cost calculation.

lmao

LiteLLM already provides this feature, no need of another proxy

i did try litellm hosted locally for privacy.

it was so bugged, the mcp page broke its postgres, many openai compatible type generic providers but not knowing what works. discovered logging was truncated by default, too late. my fault for that last one.

moved off it after the trivy incident.

bifrost

Is this a joke project?

Reply on news.ycombinator.com