Deploy Helicone

Open-source LLM observability: log cost, latency and prompts of AI calls

Deploy Helicone

Just deployed

/var/lib/clickhouse

Just deployed

/data

/var/lib/postgresql/data

Deploy and Host Helicone on Railway

Helicone is an open-source LLM observability platform. Send your OpenAI or Anthropic calls through its proxy, or log them from your code, and it records every request with cost, latency, token counts and the full prompt and response, so you can debug, monitor spend and review quality in one dashboard.

About Hosting Helicone

Self-hosted Helicone has several parts: a Next.js dashboard, the Jawn API server that also proxies LLM calls, Postgres for accounts and API keys, ClickHouse for request analytics, and S3-compatible storage for request and response bodies. The browser has to reach the dashboard, the API and the stored bodies on one origin, and the first account normally needs manual SQL because self-hosted Helicone sends no verification email. This template runs each part as its own Railway service on the private network, puts them behind one HTTPS domain, runs the database migrations on every deploy, creates your verified admin account, closes public sign-up and generates every secret.

Common Use Cases

  • Log every LLM call from your apps with cost, latency and token usage per user or feature
  • Debug agents and multi-step chains with the full request and response bodies
  • Track spend across OpenAI and Anthropic models in one dashboard
  • Keep prompts and completions in your own infrastructure for privacy or compliance

Dependencies for Helicone Hosting

  • Helicone web, jawn and migrations images, release v2025.08.21
  • PostgreSQL (Railway Postgres), ClickHouse 24.3, MinIO (S3 storage) and a Caddy router

Deployment Dependencies

Implementation Details

ServiceRoleStorage
heliconePublic router: /jawn/* to the API, bucket paths to MinIO, the rest to the dashboard–
webHelicone dashboard–
jawnAPI server and LLM proxy–
migrationsRuns Postgres and ClickHouse migrations on each deploy, creates the first admin, then exits–
clickhouseRequest analyticsVolume
minioRequest and response bodiesVolume
PostgresAccounts, organizations, API keysVolume

First login

  1. Enter your email as HELICONE_ADMIN_EMAIL when you deploy. It is the only prompt.
  2. Wait until the migrations logs end with migrations and bootstrap finished.
  3. Open https://YOUR-DOMAIN/signin and sign in with that email and the generated HELICONE_ADMIN_PASSWORD from the migrations service's variables.
  4. Create a Helicone API key under Settings, then point your client at the proxy:
    • OpenAI: base URL https://YOUR-DOMAIN/jawn/v1/gateway/oai/v1
    • Anthropic: base URL https://YOUR-DOMAIN/jawn/v1/gateway/anthropic
    • Add the header Helicone-Auth: Bearer YOUR-HELICONE-KEY. Your provider key stays in your app and is passed through.

Teammates: sign-up is closed. Set HELICONE_ALLOW_SIGNUP=true on the helicone service while teammates register, then remove it. New accounts need a verified email: add SMTP settings (SMTP_HOST, SMTP_PORT, SMTP_USER, SMTP_PASS) to web, which needs a Railway plan with outbound SMTP, or mark them verified in Postgres.

Updates: every service builds from the template repo, so Railway offers you an update when a new version is merged there. Read the changelog first. The migrations service applies schema changes on its next deploy.

Resources: about 700 MB of RAM in total when idle (measured in our test).

Why Deploy Helicone on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Helicone on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.


Template Content

More templates in this category

View Template
Rootprint
Open-source logs and traces with full-text search on object-storage.

Rootprint
1
View Template
Pyroscope profiling
Protected continuous profiling with durable Pyroscope storage.

Anton Orel
1
View Template
SigOnly
Deploy SigNoz with a working demo app & config in one click

zoeyjones
21