“Langfuse gives us complete visibility into our AI workflows. From tracking token usage and costs across multiple LLM providers to tracing every generation and embedding operation, it's become essential for understanding how our users interact with AI. The tagging and session tracking features make debugging and optimization so much easier. Thank you for building such a powerful observability platform for LLMs!”
LLM Observability & EvalsMaker says so +1 · source ↗
Langfuse
Open-source LLM engineering platform for tracing, evaluation, prompt management and experiments across the AI app lifecycle.
35 makers' products we track use Langfuse. For LLM observability & evals it's 2nd most chosen of 7 options: 35 of the 117 makers' products we know the LLM observability & evals of (30%). 58 open-source projects also declare it in their code. What makers like most: Detailed traces show each agent call and the context it pulled, which makes debugging agent behaviour much easier. Their most common complaint: Prompt management and experiments feel basic next to the tracing core. On GitHub, 9 public pull requests moved projects to Langfuse from tools we list, and 0 moved them away.
Profile checked
Works with AI agents:llms.txtCLIHow makers use Langfuse
Makers' products we track use Langfuse for LLM observability & evals (35), most often alongside Vercel (43%), OpenAI API (34%), Next.js (34%).
What Langfuse is used for
Langfuse vs the alternatives makers chose
For LLM observability & evals, 35 of the 117 makers' products we know the choice of picked Langfuse (30%), behind Datadog (43%).
| Option | Makers' products | Free for | At 10,000 users | Open source | |
|---|---|---|---|---|---|
| 50 43% | — | — | No | ||
| 35 30% | ~5,000 users | $29/mo · Core | Yes | ||
| 13 11% | ~1,000 users | $116/mo · Pro | Yes | Compare → | |
| 11 9% | ~520 users | $475/mo · Developer | No | Compare → | |
| 6 5% | — | — | No | Compare → | |
| 3 3% | — | — | No |
For LLM observability & evals: how many of the 117 makers' products we know the LLM observability & evals of chose each option, and what each costs at 10,000 users.
Costs are the cheapest plan that fits, from each vendor's pricing page, at the usage our stack planner assumes for a product with that many users. Free for: the users its cheapest plan a business can use stays free for.
Often paired with
Share of the 35 makers' products on Langfuse that show this choice.Markets that build with it most
Product Hunt markets where Langfuse is in far more of the stacks we know than across every market (1%).
When to use Langfuse, and when not
Pick Langfuse if
- You want tracing, prompts and evals in one tool you can self-host for free. LLM Observability & Evals →
Something else fits better if
- You already use OpenTelemetry or want vendor-neutral instrumentation: Arize Phoenix.
- You want request logs and costs today by changing one base URL: Helicone.
- Your app is built on LangChain or LangGraph: LangSmith.
- Your main question is whether a prompt or model change made outputs better or worse: Braintrust.
Langfuse records what your LLM app actually did on each request: the prompt that went out, the model's answer, token counts, cost, latency, and every retrieval or tool step in between. Without that record, debugging a non-deterministic agent is mostly guessing, and you have no real traffic to run evaluations against.
More about Langfuse
The core unit is a trace made of nested observations (LLM calls, spans, tool calls), grouped into sessions and users. You send them with the typed Python or JS/TS SDK, which is built on OpenTelemetry, through one of its many framework integrations (OpenAI SDK, LangChain/LangGraph, LlamaIndex, Vercel AI SDK, LiteLLM and others), or from any other language by pointing an OTLP exporter at its endpoint. The SDKs queue and batch events in the background rather than sending them inline.
On top of traces sit prompt management (versioned prompts fetched at runtime and linked to the traces they produced), datasets and experiments, scores from LLM-as-a-judge, code or human annotation, and dashboards and alerts. Langfuse Cloud runs in separate US, EU, Japan and HIPAA regions; the same codebase can be self-hosted with Docker Compose for testing or Kubernetes/Terraform for production.
Self-hosting for production means operating Postgres, ClickHouse, Redis and blob storage yourself, and a few add-on features need a license key. Cloud regions are fully separate, so switching region later means a new account and a data migration.
How Langfuse itself is built
34 tools, from its own code, website and Product Hunt page.
Langfuse pros and cons
What makers like most: Detailed traces show each agent call and the context it pulled, which makes debugging agent behaviour much easier. Their most common complaint: Prompt management and experiments feel basic next to the tracing core.
Pros
- Detailed traces show each agent call and the context it pulled, which makes debugging agent behaviour much easier. PHPH 2HN
- Token usage and cost are tracked across providers alongside quality, in one place. PHPH 2HN
- Open source and self-hostable, a common default for teams wanting tracing on their own infrastructure. PHHNHN 2HN 3
- Integrates through OpenTelemetry and native integrations in other AI tools. HNPHPH 2
Cons and watch-outs
- Prompt management and experiments feel basic next to the tracing core. HNHN 2HN 3
- Views are retrospective dashboards, so explaining one run's cost or failure still means reading trace trees by hand. HNHN 2HN 3HN 4
- The ClickHouse acquisition raised GDPR and data-residency concerns for EU users of the cloud version. HNHN 2
- Self-hosting is heavy, with ClickHouse eating disk space and several storage dependencies to run. HNHN 2HN 3
Themes that recur in makers' words and Hacker News comments, each linked to what it summarises, with how Product Hunt tags its reviews.
5.0★ from 49 reviews on Product Hunt ↗Reliability and open issues
Langfuse's status page posted 14 incidents in the last 90 days, 1 major; the most wanted open issue on GitHub has 19 👍.
Incidents on its status page
14 in the last 90 days, 1 major · 25+ in the last year- major Degraded ingestion for 20 projects Sep 2026
- minor Elevated error rates in In App Agent Sep 2026
- minor OTEL ingestion queue delay over 10 minutes in US Sep 2026
Most wanted on GitHub
- feat: render inline preview of langfuse media (images/audio/video) within trace input/output json viewer👍 19 · opened Jan 2025 · active Aug 2026
- feat: filtering sessions with metadata👍 15 · opened Mar 2025 · active Aug 2026
- Allow using `metadata` fields (e.g. `organizationId`) as Breakdown Dimension in Dashboard Widgets👍 14 · opened Mar 2026 · active Sep 2026
- bug: Filters for traces are not working like expected👍 12 · opened Feb 2026 · active Aug 2026
- bug: multi-modal support: images not showing inline in Langfuse UI👍 11 · opened Dec 2024 · active Aug 2026
Who switches
More projects move to Langfuse: 9 pull requests in, 0 out; the most come from LangSmith.
LangSmith → Langfuse9 PRs
- Migrate observability from LangSmith to Langfuse and trim logged fieldsHOSH19/HarnessLab · 2026-08-29
- feat: migrate observability from LangSmith to Langfuse Cloud (v4 SDK)icekarim/momo-assistant · 2026-08-05
- feat(trace): migrate LangSmith to Langfuse v4 and fix env key priority1935494577/Xiaoxin-Agentic-RAG-System · 2026-08-01
- [codex] Migrate from LangSmith to Langfuse, remove Python legacy, fix review issuesBunnyRabbit8mile/codex-tee · 2026-07-14
- Reapply "feat: migrate observability from LangSmith to Langfuse"sinuarlowbaby/RAG-PDF-Chatbot · 2026-07-14
- refactor: migrate observability from LangSmith to Langfusewei-yiting/fin-lab-x · 2026-03-18
- Migrate observability from LangSmith to LangfuseAneeshPulukkul/hybrid-rag-solution · 2026-03-14
- feat:migrate LLM observabilty from LangSmith to LangFuse100-hours-a-week/17-JinyUs-Q-Feed-AI · 2026-02-26
- Refactor: Migrate from Langsmith to Langfusedylangamachefl/fantasy-football-chatbot-v2 · 2025-12-01
Public pull requests on GitHub since Oct 2024 whose title says "X to Y" — real code changes moving a project from one tool to another, by developers in general.
Who uses it
35 makers' products, each linked to the source that shows it, and 58 open-source projects that declare it in their code. The best known on Product Hunt: Agentplace.io (2,313 followers), Lingo.dev (1,679 followers), Flowstep (1,620 followers), GoMarble (1,453 followers), AutonomyAI (1,366 followers). 1 of the makers' products is gone now (domain lapsed, up for sale or taken down); it chose Langfuse while live.
“Langfuse is our choice when observability and tracing are the key. Our clients love their tool and we do too!”
LLM Observability & EvalsMaker says so · source ↗“Observability for LLM apps and agents Langfuse helps us trace, debug, and improve agent quality across the platform.”
LLM Observability & EvalsMaker says so · source ↗“Detailed tracing and evals for our AI workloads.”
LLM Observability & EvalsMaker says so · source ↗“Best for LLM monitoring.”
LLM Observability & EvalsMaker says so · source ↗“Built in integration for the best LLM observability out there.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse helps us manage LLM observability, metrics, and prompts.”
LLM Observability & EvalsMaker says so · source ↗“Marc gave me an in-person onboarding in SF - I found an issue in our LLM provider config just 30 minutes after the onboarding thanks to Langfuse. 10/10 recommendation”
LLM Observability & EvalsMaker says so · source ↗“Langfuse powers our LLM observability. Without Langfuse, our AI agent would not be best-in-class. We have been using Langfuse since nearly the beginning: 2+ years!”
LLM Observability & EvalsMaker says so · source ↗“Using Langfuse as MInicule's observability layer”
LLM Observability & EvalsMaker says so · source ↗“LLM observability platform that just works!”
LLM Observability & EvalsMaker says so · source ↗“Helping us stay on top off our AI agents.”
LLM Observability & EvalsMaker says so · source ↗“needing to provide AI observability to users - natural choice.”
LLM Observability & EvalsMaker says so · source ↗“We use it to trace every model call in Flowstep, super useful for debugging”
LLM Observability & EvalsMaker says so · source ↗“Shoutout to the team at Langfuse for building a robust observability and evaluation platform for AI systems. Langfuse is a core part of Imagine.dev’s agentic architecture, giving us deep visibility into agent behavior and model performance as we iterate and scale. Their focus on tracing, metrics, and developer tooling has been critical to making our system reliable in production.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse lets us monitor the quality and cost of all our AI features with ease. And our assistant likes to use it to improve itself.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse is an amazing tool, it helps us evaluate quality and keep track on our LLM activity!”
LLM Observability & EvalsMaker says so · source ↗“Extremely relilable LLM Engineering Platform.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse helps us manage LLM observability, metrics, and prompts.”
LLM Observability & EvalsMaker says so · source ↗“We use Langfuse to keep track of our LLM prompts while building MCP-Builder.ai. It’s a great tool that makes it easy to monitor and analyze prompt performance, helping us improve quickly and efficiently.”
LLM Observability & EvalsMaker says so · source ↗“Without Langfuse, we would have been flying blind with our drafting agent. This platform is critical not just for measuring performance, but for understanding exactly what context gets pulled into our drafting agent. The key benefits of Langfuse are: 1) (Real-time) visibility into agent performance 2) Detailed tracing that shows our system's decision-making process including the context 3) Open source with self-hosting options (huge win for us) Being able to host Langfuse on our own infrastructure while getting enterprise-grade LLM observability is exactly what we needed. Highly recommended for teams serious about understanding how their LLM systems actually work and to figure out how to improve them.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse traces every LLM call Kai makes. When a summary or a draft comes out wrong, we can see exactly what the model got and fix the prompt.”
LLM Observability & EvalsMaker says so · source ↗“When you're building agentic AI workflows, observability isn't optional – it's survival. Langfuse gives us full traceability across our LLM pipelines. We can see exactly what's happening inside each agent call, catch regressions early, and continuously improve output quality. It's open-source, deeply integratable, and one of the most thoughtfully built tools in the LLM engineering space. If you're building anything with LLMs and you're not using Langfuse, you're flying blind.”
LLM Observability & EvalsMaker says so · source ↗“The ability to easily monitor, configure and debug what happens in our LLM pipelines has proved to be extremely useful during developing and QA. It's a great tool for everyone working with LLMs!”
LLM Observability & EvalsMaker says so · source ↗“Thanks for the generous free tier and excellent LLM observability 🙌”
LLM Observability & EvalsMaker says so · source ↗“Langfuse has been super helpful for us to measure the performance of our LLM interactions and making sure we're constantly improving our agents for our users.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse helps us trace every autofill decision. It's transparent, tunable, and built for trust.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse is the easiest way to add observability to your AI stack. We use Langfuse ourselves, and have a native integration for our customers to push their spans to Langfuse as well.”
LLM Observability & EvalsMaker says so · source ↗“Langfuse was the RusticAI team's first choice for tracking, debugging, and improving our AI Design Agent.”
LLM Observability & EvalsMaker says so · source ↗None of these in makers' own words.
Open source: a project that declares Langfuse as a dependency in its public code — verifiable, but not necessarily a live product.
What its projects tell coding agents
29 of the 45 open-source products built with Langfuse that were worked on in the last 180 days keep instruction files for coding agents — 64%, against 53% of all the projects we read. AGENTS.md is in 24, CLAUDE.md in 24.
Alternatives to Langfuse
All alternatives by situation →Questions makers ask about Langfuse
Is Langfuse good for production?
35 makers' products we track run on it, and 1 of them have since gone (the product, not necessarily because of Langfuse). Its own status page posted 14 incidents in the last 90 days, 1 of them major; vendors decide what they post there, so this doesn't compare across tools. On GitHub, 9 public pull requests moved projects to Langfuse from tools we list, and 0 moved them away. The watch-out makers raise most: Prompt management and experiments feel basic next to the tracing core.
What is Langfuse used for?
Open-source LLM engineering platform for tracing, evaluation, prompt management and experiments across the AI app lifecycle. Makers' products we track use it for LLM observability & evals (35).
What are the pros and cons of Langfuse?
Pros makers name: Detailed traces show each agent call and the context it pulled, which makes debugging agent behaviour much easier. Token usage and cost are tracked across providers alongside quality, in one place. Open source and self-hostable, a common default for teams wanting tracing on their own infrastructure. Cons: Prompt management and experiments feel basic next to the tracing core. Views are retrospective dashboards, so explaining one run's cost or failure still means reading trace trees by hand. The ClickHouse acquisition raised GDPR and data-residency concerns for EU users of the cloud version.
What are the alternatives to Langfuse?
For LLM observability & evals, the other options makers' products we track chose most are Datadog (50), Helicone (13), LangSmith (11), Braintrust (6), Weights & Biases (3). Open source or self-hostable among them: Helicone, Weights & Biases. Each alternative, with the situation it fits, is on the Langfuse alternatives page.
How much does Langfuse cost per month?
At 10,000 users, Langfuse costs $29 a month on the cheapest plan that fits (Core), from its pricing page.
Do developers switch to or away from Langfuse?
On GitHub, 9 public pull requests moved projects to Langfuse from tools we list, most from LangSmith (9); 0 moved them away.
When should you use Langfuse?
You want tracing, prompts and evals in one tool you can self-host for free (LLM observability & evals). Something else fits better if: you already use OpenTelemetry or want vendor-neutral instrumentation, pick Arize Phoenix; you want request logs and costs today by changing one base URL, pick Helicone.
Where is my data stored on Langfuse Cloud?
You pick a region at sign-up, US (Oregon), EU (Ireland), Japan (Tokyo) or a dedicated HIPAA region, all on AWS. Regions share no accounts or data, so moving later means creating a new account and migrating. source ↗
Is it SOC 2 compliant, and can I use it with health data?
Langfuse Cloud is covered by SOC 2 Type II and ISO 27001 audits and offers a DPA for GDPR. For PHI there is a separate HIPAA-ready region where customers on the Pro plan or higher can sign a BAA. source ↗
What does self-hosting involve?
You run the same containers as Langfuse Cloud (a web and a worker container) plus Postgres, Redis or Valkey, ClickHouse and S3-compatible storage. Docker Compose is meant for local use and testing; for production they provide Helm and Terraform for AWS, Azure and GCP. source ↗
Can I send traces from a language other than Python or TypeScript?
Yes. Langfuse accepts OpenTelemetry traces on its OTLP endpoint, so any language with an OpenTelemetry SDK, or OTel-based instrumentation libraries, can export spans to it. source ↗
Does tracing add latency to my requests?
The SDKs and integrations queue events and send them in batches in the background, based on batch size and a flush interval. In short-lived processes such as serverless functions you call flush before exiting so queued events are not lost. source ↗
How long is trace data kept?
Each Cloud plan has a data access window (30 days on Hobby). Custom per-project retention, with a minimum of 3 days, is available on paid plans; without a policy, stored data is not deleted automatically. source ↗
Can I move between self-hosted and Cloud later?
Yes. Langfuse publishes a Python migration cookbook that copies traces, prompts and datasets between a self-hosted instance and Langfuse Cloud, or between two instances. source ↗
Is Langfuse free?
Yes — there is a free tier a small product can run on; paid use starts at $29/mo. source ↗
Is Langfuse open source or self-hostable?
Open source, and you can self-host it. source ↗
Can AI coding agents work with Langfuse?
It serves an llms.txt docs index; it has an official CLI (langfuse-cli).
Who uses Langfuse?
35 makers' products we track, each with a source, and 58 open-source projects declare it in their code. The best known of them on Product Hunt: Agentplace.io (2,313 followers), Lingo.dev (1,679 followers), Flowstep (1,620 followers), GoMarble (1,453 followers), AutonomyAI (1,366 followers). source ↗