详细内容
Backed by
Open Source LLM Engineering Platform
Traces, evals, prompt management and metrics to debug and improve your LLM application.
Auto-advance isactive. Press Escape to pauseauto-advance.
Observability
Metrics
Prompt Management
Playground
Evaluation
Annotations
Public API
Python SDK
JS/TS SDK
from langfuse import observe
# drop-in wrapper adds OpenTelemetry tracing to OpenAI
# many other llm/agent integrations are available
from langfuse.openai import openai
@observe() # decorate any function; all nested calls are auto-linked
def handle_request(text: str) -> str:
res = openai.chat.completions.create(
model="gpt-5",
messages=[\
{"role": "system", "content": "Summarize in one sentence."},\
{"role": "user", "content": text},\
],
)
return res.choices[0].message.content
from langfuse import observe
# drop-in wrapper adds OpenTelemetry tracing to OpenAI
# many other llm/agent integrations are available
from langfuse.openai import openai
@observe() # decorate any function; all nested calls are auto-linked
def handle_request(text: str) -> str:
res = openai.chat.completions.create(
model="gpt-5",
messages=[\
{"role": "system", "content": "Summarize in one sentence."},\
{"role": "user", "content": text},\
],
)
return res.choices[0].message.content
Observability
Capture complete traces of your LLM applications/agents. Use traces to inspect failures and build eval datasets. Based on OpenTelemetry with support for all popular LLM/agent libraries.
Teams building complex LLM apps rely on Langfuse
\
\
Read StoryRead Customer Story 
\
\
Read StoryRead Customer Story 
\
\
Read StoryRead Customer Story 
\
\
Read StoryRead Customer Story
13,604,327+
SDK installs / month
17,236
GitHub stars
5,515,268+
Docker pulls
Works with any LLM app and model
SDKs for Python & JS/TS, native integrations for popular libraries and support for OpenTelemetry
Integration overview Request new integration
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\Proudly Open Source
Langfuse is committed to open source. You can also run it locally or self-hosted.
Changelog
Hosted MCP Server for Langfuse Prompt Management by Michael\
11 days ago OpenAI GPT-5.1 support by Hassieb\
Nov 14, 2025 Organize Your Datasets in Folders by Marlies\
Nov 8, 2025 JSON Schema Enforcement for Dataset Items by Hassieb\
Nov 8, 2025 Score Analytics with Multi-Score Comparison by Michael\
Nov 7, 2025 Annotation Support in Experiment Compare View by Marlies\
Nov 6, 2025 Baseline Support in Experiment Compare View by Marlies\
Nov 6, 2025 Filters in Compare View by Marlies\
Nov 6, 2025 Langfuse for Agents by Nimar\
Nov 5, 2025 Amazon Bedrock AgentCore Integration by Marc\
Nov 4, 2025 @Mentions and Reactions in Comments by Michael\
Nov 4, 2025 IdP-Initiated SSO Support by Marc\
Nov 4, 2025 Mixpanel integration by Marc\
Nov 4, 2025 Advanced Filtering for Public Traces and Observations API by Valeriy\
Nov 3, 2025 Filter Sidebar for Tables by Nimar, Leo\
Nov 3, 2025 Langchain v1 Support by Hassieb\
Oct 26, 2025 LLM-as-a-Judge Execution Tracing & Enhanced Observability by Hassieb\
Oct 16, 2025 Spend Alerts by Michael\
Oct 10, 2025 Natural Language Filtering for Traces by Nimar, Marlies, Leo\
Sep 30, 2025 Structured Output Support for Prompt Experiments by Hassieb\
Sep 30, 2025
Latest Release
v3.135.1
21 hours ago
Self-Host Langfuse
🚀
⚡
DockerDocker Compose Kubernetes (Helm) AWSAWS (Terraform) GCPGCP (Terraform) AzureAzure (Terraform)
Community Q&A
Missing input_tokens and output_tokens fields in the Metrics API (only total_tokens available)
@JimHeo
1↑1💬
llm-as-a-judge evaluation didn't correctly fetch data
@JoeyWangTW
1↑2💬
Input Output shown not support Thai language
@wyotthawee
1↑2💬
Is there a way to filter traces that have called an llm?
@ThePalo
1↑1💬
Is there a way to see the trace of a LLM-as-judge evaluation? the actual input or output?
@mfcabrera
✓1↑2💬
Trying to call the langfuse batch ingestion api from a lambda function, all events not visible on dashboard despite 201 success
@rjayaram-godaddy
✓1↑13💬
Connecting to Langfuse API when self-hosting and using Google GCP IAP
@ayoubhamaoui
1↑4💬
Routing @observe() decorator traces to external OTLP endpoint (Grafana Alloy/Tempo)
@mscherrmann
✓1↑2💬
LLM-as-a-judge on observations instead of traces
@thiagosalvatore
✓2↑4💬
Add tags to traces post creation via the Python SDK
@anshmove37
✓1↑2💬
mem0 integration
@sacredvoid
✓1↑1💬
“Body exceeded 1mb limit” error on dataset upload
@kourosh-samia
1↑2💬
How to add expected output for Geval?
@divyamuralidharan-ai
✓1↑1💬
Is it possible to plot latency graph for each agents in langfuse dashboard.
@divyamuralidharan-ai
1↑2💬
Breaking change in @langfuse/langchain v4 beta 2
@ghaliela
1↑3💬
Recommended way to flush logs in AWS Lambda using Vercel AI SDK + @opentelemetry/sdk-node
@kylegwalsh
✓3↑1💬
[Alerts] Connect to Slack
@hypernovas
✓10↑2💬
BUG: The "OUTPUT" is not being displayed in "formatted" style
@mathpinh
2↑2💬
can i ingest llm cost via api
@ZMRWEGO
✓1↑4💬
Invalid remote run URL:Blocked IP address detected:10.11.12.250
@cantyK
✓1↑3💬
1,284
threads
Roadmap Discussions
WordPress.com SSO configuration option
@darthhexx
1↑1💬
Custom scopes for public api keys
@marcklingen
5↑4💬
Add Causal Chain Visualization for LLM Trace Debugging
@Aki-07
1↑0💬
Allow getting evaluator ID from score Object created by LLM-as-a-Judge evaluator
@erantr1
1↑2💬
Ability to tag a single prompt version
@elyas-bhy
1↑1💬
Format tool definitions in observations
@verdie-g
1↑0💬
Provide renaming functionality for datasets in Langfuse public API
@vserecun
1↑0💬
will be support langchain v1?
@miniboxhaha
1↑2💬
Support Dynamic Headers in Otel Traces and Langfuse API calls
@tw-dpd
2↑2💬
Feature Request: Option to Export Annotated Score Values and Comments from Human Annotation Queues
@Neha-Jacob
1↑0💬
Option to display costs in other currencies
@jannikmaierhoefer
2↑1💬
return prompt type in list prompt API
@xxchan
1↑2💬
Support Ollama API in langfuse playground (and LLM API / Gateway)
@LiveOverflow
1↑2💬
Structured Output, on prompts in prompt mgmt and independent schema management
@borisevstratov
17↑7💬
Ability to trace calls through MCP with its associated tools
@ansu86d
5↑2💬
Export a trace in OTEL format so it can be viewed in tools like `otel-tui`
@anton-pt
3↑1💬
Ability to view reasoning messages in prompt playground
@6matt
1↑1💬
Support setting custom model usage prices via API/SDK (parity with UI)
@ikazoy
4↑2💬
Open playground in new tab
@marctorsoc
2↑1💬
Better A/B testing with random labels
@danrex
2↑1💬
1,131
threads
Enterprise Security
Langfuse is built with security in mind. Langfuse Cloud is SOC 2 Type II and ISO 27001 certified, GDPR compliant and aligned with HIPAA.
Customer Stories
See what our customers are saying about Langfuse
[
\
\
"Langfuse has enabled our developers to get extremely fast feedback. It's fundamental to how our developers understand their AI implementations."\
\
\
\
Walt Wells\
\
Staff Software Engineer at Khan Academy](https://langfuse.com/customers/khan-academy)
[
\
\
"When a customer pings us, we pop open Langfuse to understand what's going on. It keeps customer support resolutions at less than 8 minutes on average."\
\
\
\
Alexander Danilowicz\
\
Co-founder at Magic Patterns](https://langfuse.com/customers/magic-patterns-ai-design-tools)
[
\
\
"Langfuse enables us to track every prompt, response, cost, and latency in real time, turning black-box models into auditable, optimizable assets."\
\
\
\
Walid Mehanna\
\
Chief Data & AI Officer at Merck](https://langfuse.com/customers/merckgroup)
[
\
\
"Building on Langfuse we saved 30% of external BPO cost by deflecting 50% of support conversations to AI."\
\
\
\
Ana Casado\
\
Head of Operations Data and AI at SumUp](https://langfuse.com/customers/sumup)
[\
\
"Langfuse has enabled our developers to get extremely fast feedback. It's fundamental to how our developers understand their AI implementations."\
\
\
Walt Wells\
\
Staff Software Engineer at Khan Academy](https://langfuse.com/customers/khan-academy)
[\
\
"When a customer pings us, we pop open Langfuse to understand what's going on. It keeps customer support resolutions at less than 8 minutes on average."\
\
\
Alexander Danilowicz\
\
Co-founder at Magic Patterns](https://langfuse.com/customers/magic-patterns-ai-design-tools)
[\
\
"Langfuse enables us to track every prompt, response, cost, and latency in real time, turning black-box models into auditable, optimizable assets."\
\
\
Walid Mehanna\
\
Chief Data & AI Officer at Merck](https://langfuse.com/customers/merckgroup)
[\
\
"Building on Langfuse we saved 30% of external BPO cost by deflecting 50% of support conversations to AI."\
\
\
Ana Casado\
\
Head of Operations Data and AI at SumUp](https://langfuse.com/customers/sumup)
[\
\
"Langfuse has enabled our developers to get extremely fast feedback. It's fundamental to how our developers understand their AI implementations."\
\
\
Walt Wells\
\
Staff Software Engineer at Khan Academy](https://langfuse.com/customers/khan-academy)
[\
\
"When a customer pings us, we pop open Langfuse to understand what's going on. It keeps customer support resolutions at less than 8 minutes on average."\
\
\
Alexander Danilowicz\
\
Co-founder at Magic Patterns](https://langfuse.com/customers/magic-patterns-ai-design-tools)
[\
\
"Langfuse enables us to track every prompt, response, cost, and latency in real time, turning black-box models into auditable, optimizable assets."\
\
\
Walid Mehanna\
\
Chief Data & AI Officer at Merck](https://langfuse.com/customers/merckgroup)
[\
\
"Building on Langfuse we saved 30% of external BPO cost by deflecting 50% of support conversations to AI."\
\
\
Ana Casado\
\
Head of Operations Data and AI at SumUp](https://langfuse.com/customers/sumup)
Previous slideNext slide
Pricing
Get started on the Hobby plan for free. No credit card required.
Langfuse Cloud (we host)Self-hosted (you host)
Hobby
Get started, no credit card required. Great for hobby projects and POCs.
Free
Trusted by:
40,000+ builders
- All platform features (with limits)
- 50k units / month included
- 30 days data access
- 2 users
- Community support via GitHub
Unlimited Users
Core
For production projects. Longer data retention and unlimited users.
$29/ month
Trusted by:
- Everything in Hobby
- 100k units / month included, additional: $8/100k units. Lower with volume ( pricing calculator)
- 90 days data access
- Unlimited users
- In-app support
Unlimited Users
Pro
For scaling projects. Unlimited history, high rate limits, all features.
$199/ month
Trusted by:
- Everything in Core
- 100k units / month included, additional: $8/100k units. Lower with volume ( pricing calculator)
- Unlimited data access
- Data retention management
- Unlimited annotation queues
- High rate limits
- SOC2 & ISO27001 reports, BAA available (HIPAA)
- Prioritized in-app support
+ optional
Teams Add-on$300/mo
- Enterprise SSO (e.g. Okta)
- SSO enforcement
- Fine-grained RBAC
- Support via Dedicated Slack Channel
Enterprise
For large scale teams. Enterprise-grade support and security.
$2499/ month
Trusted by:
- Everything in Pro + Teams
- 100k units / month included, additional: $8/100k units. Lower with volume ( pricing calculator)
- Audit Logs
- SCIM API
- Custom rate limits
- Uptime SLA
- Support SLA
- Dedicated support engineer
+ optional
Yearly Commitment
- Custom Volume Pricing
- Custom Terms & DPA
- Architecture reviews
- Billing via AWS Marketplace
- Billing via Invoice
- Vendor Onboarding
For a detailed comparison and FAQ, see our pricing page.
Join the Community
80.000+ people have built on Langfuse.
Follow @langfuse Share some love
\
\
\
\
Manu Hortet\
\
Founder at Producta\
\
Complete and simple, sweet product. We also love it is open source! 
\
\
\
\
Jorge Torres\
\
CEO @ MindsDB\
\
One of my fav products! Long Live Langfuse, we use it everywhere in MindsDB 
\
\
\
\
Nadeesha Cabral\
\
SVP of Engineering\
\
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.
\
\
\
\
Boris Arzentar\
\
Co-founder of Cognee\
\
Langfuse is our choice when observability and tracing are the key. Our clients love their tool and we do too! 
\
\
\
\
Viktor Sarstrom\
\
@viktorsarstrom
@langfuse is goat llm observability 🐐
\
\
\
\
\
Dave Wilson\
\
Supabase\
\
Brilliant product, this team is smashing it! 
\
\
\
\
Othmane Zoheir\
\
@othmanezoheir
I've been using @langfuse from day 1 for my LLM package "tinyllm" and all my LLM apps. It just works and the team is fast and very helpful. Probably the best open-source tool/experience I've ever had.
\Show more
Start building with Langfuse today.
Choose your path to start building better LLM applications
Integrate into your application
Integrate Observability (15 min) \
\
Trace and monitor your LLM applications/agents Integrate Prompt Management (5 min) \
\
Version and optimize your prompts collaboratively Run Evaluations (15-30 min) \
\
Test and improve your LLM outputs with online/offline evals
Dive deeper
Explore Example Project (1 min) \
\
Click around and test all features in a live demo project Watch Walkthroughs (20 min) \
\
Get to know all platform areas via pre-recorded demo Read Documentation \
\
Technical docs, examples, and guides to understand all details of the platform