Langfuse hosting
Tracing, evaluation and prompt management for LLM applications.
Langfuse hosting at AIHostBox means a dedicated, single-tenant Langfuse instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $11.99 per month and the admin account, the data and any API keys are yours.
Langfuse is an open-source LLM engineering platform. Your application sends traces through the Python or JavaScript SDKs, through integrations for OpenAI, LangChain, LlamaIndex and LiteLLM, or through an OpenTelemetry endpoint, and Langfuse stores each call with its inputs, outputs, latency, token usage and cost. Around that sit prompt management with versions and labels, datasets, experiments and LLM-as-a-judge evaluations, which is why teams use it both for debugging and for deciding whether a change actually made things better.
Unlike a lightweight logger, Langfuse is a small data platform. Version 4 runs a web container and a worker, with Postgres for users and prompts, ClickHouse for traces, observations and scores, Redis for queues, and S3-compatible storage where every incoming event is written first. That is five moving parts to keep patched and backed up. The core, including tracing, prompts and evaluations, is MIT-licensed; audit logs, data retention policies and project-level RBAC need an enterprise licence key.
Langfuse hosting plans
Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.
How we run Langfuse
Langfuse’s official Docker Compose stack: langfuse-web and langfuse-worker (v4 images), Postgres, ClickHouse, Redis and MinIO, with only the web UI and API published over TLS.
What we configure for you
- NEXTAUTH_SECRET, SALT and ENCRYPTION_KEY generated as long random values instead of the CHANGEME defaults
- NEXTAUTH_URL set to your hostname so links and OAuth callbacks resolve
- CLICKHOUSE_PASSWORD, REDIS_AUTH and the MinIO keys behind LANGFUSE_S3_EVENT_UPLOAD_* and LANGFUSE_S3_MEDIA_UPLOAD_* changed from defaults
- SMTP_CONNECTION_URL and EMAIL_FROM_ADDRESS for invitations and password resets
- SSO through AUTH_<PROVIDER>_* variables, plus AUTH_DISABLE_SIGNUP for a closed instance
- TELEMETRY_ENABLED=false if you prefer no usage telemetry
Which plan fits
Plus (4 GB, 2 vCPU) is enough for a developer or small team tracing a prototype or staging environment. Pro (8 GB, 3 vCPU) gives ClickHouse room for production apps with steady but moderate traffic. Both sit well under Langfuse’s own 16 GiB single-VM guidance, so very high ingest belongs on a larger cluster.
The honest limits
Langfuse notes that its Docker Compose setup has no high availability or horizontal scaling, so an instance scales only vertically; enterprise features such as audit logs and retention policies need a paid licence key.
What people run Langfuse for
Debug agent behaviour
Follow a single request through every tool call.
Prompt regression testing
Run a new prompt against a saved dataset before shipping it.
Cost attribution
See which feature is generating the spend.
Licence and version policy
| Item | Position |
|---|---|
| Upstream licence | MIT (core) |
| Sold as managed hosting | Yes |
| Tenancy | Single tenant — one customer per instance |
| Patching | Critical advisories within 24 hours, high severity within 72 |
Our full position on licences, including the applications we decline to sell, is on the licensing page.
Reading on Langfuse
Works well with Langfuse
More Gateways & LLM ops we host
Langfuse hosting: your questions
Which parts of Langfuse are open source?
Can I sign in with Google, Okta or Entra ID?
Where is my trace data stored, and what is backed up?
Do I need to change my code to send traces here?
How much traffic can these plans handle?
Run Langfuse without running a server
A dedicated Langfuse instance, deployed and maintained by us. Month to month, cancel whenever.