Langfuse on Railway
Open-source LLM observability: traces, evaluations and prompt management, self-hosted on Railway.
Langfuse is an open-source platform for observing LLM applications. It collects traces of model calls and agent runs, tracks cost and latency, supports evaluations and datasets, and manages prompt versions. It is often compared with LangSmith and Helicone, with the difference that you can run it on your own infrastructure.
Teams self-host Langfuse to keep prompts and model outputs in their own environment. Version 4 stores traces in ClickHouse, application data in Postgres, queues in Redis and raw events in S3-compatible storage, and accepts data through its SDKs or OpenTelemetry.
The Langfuse core is MIT-licensed. Code in its ee directories uses the Langfuse enterprise licence.
What Langfuse can do
- Tracing for LLM calls and agent runs
- Cost and latency tracking
- Evaluations, datasets and experiments
- Prompt management with versions
- OpenTelemetry endpoint for SDKs and exporters
- Projects, organisations and API keys
What people use it for
- Debugging and tracing an LLM application or agent
- Tracking token cost across models and users
- Versioning and testing prompts
- Running evaluations on production traces
Videos
Running Langfuse on Railway
The Railway template deploys the full Langfuse v4 stack: the web app and a background worker pinned to 4.56.0, ClickHouse, Postgres, Redis and a Railway Bucket for raw events, media and exports. The first admin, organisation, project and API keys are created on first start and public sign-up is disabled. Cost follows memory across the five services plus volumes and bucket storage; the web service needs at least 1.5 GB, and ClickHouse is the other main consumer as ingestion grows.
Langfuse templates on Railway
Tracklayer templates are listed first. Other Langfuse templates follow, ordered by total deployments on Railway. How this works.
Deployment numbers as of 2026-11-20.
Questions about Langfuse
How do I log in?
Enter your email in LANGFUSE_INIT_USER_EMAIL when deploying. After langfuse-web is healthy, which can take several minutes on first start while migrations run, sign in with that email and the generated LANGFUSE_INIT_USER_PASSWORD from the langfuse-web variables, then change the password.
Which Railway plan do I need?
Hobby or Pro. langfuse-web needs at least 1.5 GB of memory, so Trial (1 GB per service) and Free (0.5 GB) cannot run it; it stops with a JavaScript heap out of memory error.
Is my data kept across redeploys?
Yes. ClickHouse, Postgres and Redis each use a volume, and raw events, media and exports go to the Railway Bucket.
How do I upgrade?
Web and worker are pinned to the same version and must be upgraded together. Change the image tag in the web Dockerfile and on the worker, check the Langfuse changelog and redeploy. Migrations run when the web service starts.
What is the licence?
The core is MIT. Code under the ee directories uses the Langfuse enterprise licence and is not enabled by this template. ClickHouse is Apache-2.0.
Guides
Similar apps
Tracklayer maintains the templates marked as such; Langfuse is a trademark of its owners and this site is not affiliated with them.