Is Langfuse still being developed after joining ClickHouse?
Langfuse says its cloud "keeps running as-is" and that it stays open source and self-hostable, with no planned licensing changes.
Resources / Compare / Langfuse alternative
COMPARE · LANGFUSE
Langfuse is a full LLM engineering platform: tracing, evals, prompt management. If what you actually open it for is the bill, LLMtrack is the smaller tool. It does AI cost tracking per feature and customer, compares each customer's cost with what they pay you, and stores no prompts.
Quick answer
Keep Langfuse if you use its traces, evals or prompt management. Pick LLMtrack if you only need costs and want margin per customer, a forecast with your budget day, and model-switch advice, with nothing stored but token counts. Plenty of teams run both.
Langfuse breaks cost down by model, tags and users, and it prices usage types such as cached, audio and reasoning tokens separately. It's open source and self-hostable, and after joining ClickHouse it says "Langfuse Cloud keeps running as-is".
So the case for LLMtrack isn't missing basics. It's scope: a tool built only for the business side of the bill.
Sources: Langfuse cost tracking docs · Langfuse joins ClickHouse
| Langfuse | LLMtrack | |
|---|---|---|
| Built for | Tracing, evals, prompt management, cost | Cost, margin and forecasting only |
| Cost by user and tag | Yes | Yes, by feature, model and customer |
| Customer cost against plan price | Not in its cost tracking docs | Yes, margin per account |
| Forecast with budget breach day | Not in its cost tracking docs | Yes |
| Model-switch advice | Not in its cost tracking docs | Switch, Test first or Keep, from independent benchmarks |
| What gets stored | Full traces, built to include inputs and outputs | Token counts and cost. Prompts are never sent |
| Open source, self-host | Yes | No. SDKs are open source; the service is hosted |
| Free plan | 50k units/mo · 30 days · 2 users | 3,000 requests/mo · 30 days · every feature |
| Paid plans | Core $29/mo · Pro $199/mo | Pro $11.99/mo · 25,000 requests · 90 days |
Langfuse figures from langfuse.com/pricing and its docs, 16 Oct 2026. Langfuse meters "units"; LLMtrack counts requests, so the free tiers aren't directly comparable.
LLMTRACK FITS IF
STAY ON LANGFUSE IF
Running both is common. Langfuse answers what a chain of calls did; LLMtrack answers what it cost and whether that customer is profitable. The guide to cost per customer for AI features shows the second question worked through.
01
Install the SDK
The LLMtrack SDK for Node and Python is one install and one call after each response. Your Langfuse tracing can stay as it is.
pip install llmtrack-sdk
from llmtrack_sdk import LLMtrack, from_openai_usage
llmtrack = LLMtrack(api_key=os.environ["LLMTRACK_API_KEY"])
res = client.chat.completions.create(model=model, messages=messages)
llmtrack.track(provider="openai", model=res.model,
feature="summarizer", customer_id="acme-corp",
**from_openai_usage(res.usage))02
Reuse your Langfuse names
Use the same tag you already give each call in Langfuse as the LLMtrack feature, and your Langfuse user ID as customer_id. The two dashboards then line up.
03
Add plan prices
Enter what each plan charges, and every customer shows a margin. Check a feature's cost before it ships with the LLM cost calculator.
Is Langfuse still being developed after joining ClickHouse?
Langfuse says its cloud "keeps running as-is" and that it stays open source and self-hostable, with no planned licensing changes.
Can LLMtrack show traces?
No. It records token counts, model, feature, customer and cost for each request. For traces, keep Langfuse or LangSmith.
Where do LLMtrack's prices come from?
Each provider's published schedule, re-synced daily. The full LLM API pricing sheet is public and dated.
Does it work with LiteLLM or OpenRouter?
Yes. Paste one URL into the gateway's trace settings. See all integrations.
Free plan with every feature. No card.
Sources, checked 16 Oct 2026:langfuse.com/pricinglangfuse.com, token and cost trackinglangfuse.com/blog/joining-clickhouse