PromptSpend

Congrats on the launch

Finally, someone addressed the biggest issue with these calculators. Most of them are outdated within a month. Factoring in compounding chat history and reasoning tokens is a huge differentiator—those are always the silent budget killers.

Love this idea! Keeping track of LLM costs across different models can be challenging. PromptSpend makes that process much more transparent and manageable. Great work, and congratulations on the launch! 🚀

Showing the source URL and check date on every price is what makes this trustworthy. Most calculators I've used were quietly months out of date. When two vendor pages disagree, does the API return both prices or just a flag?

I like that the LLM price list cites its sources, since pricing pages change constantly and stale numbers are a common problem. Would a changelog showing when each price last changed be on the roadmap?

I like that every price carries its source URL and confirmation date, and that disagreeing vendor pages get flagged instead of averaged. The handling of compounding chat history in the cost estimate is something most calculators skip.

I built PromptSpend because LLM cost calculators go stale the moment a vendor reprices. Every price here shows its source and the date it was last confirmed. Feedback very welcome!

I review AI tools and plans for a living, and stale pricing is the thing I fight most, so a source URL and check date on every number is exactly right. Question: when a vendor quietly changes a price mid-month, do you keep the old figure visible anywhere? Being able to say "this went up on the 12th" would make it useful for writing about pricing changes, not just budgeting.

Showing the source URL and confirmation date on every price is the part I trust most here. How do you handle vendors that change prices mid-day or show different prices by region? Do you keep a history so people can see when a model got cheaper?

Showing the source URL and confirmation date on every price is what makes this trustworthy — most calculators go stale the moment a vendor reprices, and flagging disagreements instead of averaging over them is the right call. Counting compounding chat history and hidden reasoning tokens is also the part other tools miss; those two quietly dominate real bills. Two questions: when vendors disagree on a price, how do you decide which figure the estimate uses, and does the comparison separate cached-input pricing from regular input pricing? The keyless API and MCP server would be useful for pulling current prices into my own loop-building workflow.

Cost tracking for LLM APIs is essential but fragmented. Most teams manually track Claude vs GPT vs Gemini pricing across tabs. Having verified sources for all pricing changes beats copy-paste spreadsheets. This is the reference guide every startup building on LLMs actually needs.

I ship LLM features at work, and hidden reasoning tokens are exactly where our early estimates were furthest off, so that line alone makes this useful. Two things I'd use: a toggle for batch API pricing (often half of list) and a cache hit rate slider, since prompt caching changes the monthly total a lot when the system prompt is long.

source url and confirmation date on every price is the right call, stale pricing is worse than no pricing. counting hidden reasoning tokens is a nice touch too, those quietly double a bill. does the api let you diff a model's price over time?

Pricing with a source link and a confirmation date is the detail most comparison tables skip — and it's the one that matters when you're actually budgeting API spend. Counting cumulative conversation history and hidden reasoning tokens is spot on too; those are the two costs people consistently underestimate. Curious how you estimate the hidden reasoning tokens when vendors don't disclose them, and how the MCP server holds up in day-to-day use?