Guides / culpa vs langfuse
Culpa vs Langfuse
Langfuse is an MIT-licensed LLM engineering platform you can self-host on your own infrastructure, and it has joined ClickHouse. Culpa, a local-first LLM cost, margin, and forecast ledger, is neither more open nor more private than that. What Culpa adds is margin against the revenue a customer pays you, and a forecast of the next month.
Why this happens
Langfuse is the strongest privacy and portability answer in this category, and pretending otherwise would be dishonest. It's MIT licensed, it ships Docker, Kubernetes and Terraform paths for AWS, GCP and Azure, and it says outright that it won't lock in your data. Culpa can't win that argument and doesn't try. Langfuse also announced it has joined ClickHouse, which matters less for what the product does today than for where its roadmap points, and that's worth asking about rather than guessing at. The separation is scope: an engineering platform answers what your calls did, and a ledger answers what they were worth.
What this usually looks like
- You have full traces and still can't rank customers by profitability.
- Cost per feature exists and cost per feature against revenue doesn't.
- Your forecast for next month is last month, unadjusted.
- A tool choice is being made on licence terms rather than on the question you have.
Free, no card, no account
Run the free Cost Leak Scan
It shows your most expensive conversation before you install anything.
Mistakes that cost the most
| Mistake | Why it hurts | Do instead |
|---|---|---|
| Assuming a commercial tool must be less open than an open-source one. | It usually is, and here Langfuse is genuinely more portable. Pretending otherwise loses trust. | Pick on the question you need answered, and treat licence and hosting as separate axes. |
| Ignoring an acquisition because the product still works. | Langfuse says its goal is unchanged, and a roadmap is still worth asking about rather than assuming. | Ask where the product goes now, the same way you would of any vendor that changed hands. |
| Treating cost attribution as the finish line. | Attribution says where the money went. It never says whether the money was worth spending. | Add revenue. Margin per customer is the number that changes a pricing decision. |
Run this check tonight
- List what you need answered, then mark which answers need revenue in the same view.
- Ask your current tool for margin on your three largest accounts.
- Ask it for next month as a range, from your own history.
- Separately, decide how much licence portability is worth to you, because it's a real axis.
- Anything left unanswered by both is the actual gap.
Where Langfuse wins, and where the question changes
Langfuse's own site and pricing, read 2026-08-03, against what Culpa does. No pricing comparison is drawn: Langfuse prices per unit of ingestion and Culpa per tracked call, so a like-for-like number would need an assumed traffic mix.
Langfuse wins the openness argument outright and the privacy argument is a draw. What's left is whether you need the finance half of the question, and that's the only ground where these two genuinely differ.
Every number, with its confidence and source
| Figure | What it means | Confidence | Source |
|---|---|---|---|
| $29 to $199 per month | Langfuse's published Core and Pro cloud plan prices, before usage above the included units | provider-reported | langfuse.com/pricing, read 2026-08-03. Both include 100k units a month and charge $8 per additional 100k. Data access is where they differ: 90 days on Core and 3 years on Pro, per the same feature table. Pro's own blurb says unlimited history, which its table contradicts, so the table is what is reported here. A Hobby tier is free at 50k units and Enterprise is listed above Pro. Self-hosting is separately available under the MIT licence. |
What a generic answer can’t know
No comparison decides whether you need margin and a forecast, because that depends on whether anyone is asking you if a customer is profitable. When they do, cost alone can't answer. Culpa joins it to revenue on your infrastructure, keeps your prompts and responses there, and counts the calls to run your plan.
Questions founders ask next
Is Culpa open source like Langfuse?
No. Langfuse is MIT licensed and self-hostable with Docker, Kubernetes and Terraform paths, and it states plainly that it won't lock in your data. That's a genuine advantage and it belongs on this page rather than buried.
Does the ClickHouse acquisition change anything?
Not for what the product does today, by its own account. Langfuse announced it has joined ClickHouse and says the goal is unchanged. The fair question is where a roadmap points after an acquisition, and that's worth asking the vendor rather than inferring.
So why choose Culpa?
For the finance half. Margin per customer against the revenue they pay you, and a forecast of next month from your own history. Checked against Langfuse's own pages on 2026-08-03, it claims neither. The same check across eleven other tools in this category found none of them claiming either, though that's a statement about the twelve that were read rather than about every tool that exists.
On your infrastructure
Culpa runs on your infrastructure. Your prompts and responses never leave it. Culpa counts calls to run your plan, and it fails open, so if it ever breaks your app keeps running.
How Culpa works
Find the culprit. Not just the total.
Your dashboard shows what you spent. It stops short of who spent it. Culpa shows the conversation, the user and the feature behind it.
Your prompts stay local.
Culpa runs on your own infrastructure. What you send to a model reaches us at no point.
Every dollar has a name.
Follow any charge to the conversation, the user, the feature and the customer behind it.
See the bill before it lands.
Cost your next feature before you ship it. You get the likely bill and the worst case, at best, median, p90 and p99.
Three steps to your first answer.
Change one base URL.
Or drop in the Python or TypeScript library.
Find your most expensive conversation.
In the first session, not the first week.
Cost your next feature before you ship it.
Why the bill went up
Example dashboardCalls traced
418,209
across 3 projects
Spend this week
$378.41
+ $182 vs last week
Failed calls
312
74% retried, and you paid for all of them
+ $182 this week traced to one culprit
Spend over 14 days
Most expensive users
Next week forecast
Graded against reality. Accuracy shown as results land.
Free, no card, no account
Run the free Cost Leak Scan
It shows your most expensive conversation before you install anything.
Keep reading
Sources: Langfuse, Langfuse pricing, ClickHouse acquires Langfuse. Last reviewed 2026-08-03. Plain text version.