--- title: "Deploy 9Router" description: "9Router AI gateway: one OpenAI-compatible API for 60+ LLM providers" category: "AI/ML" url: https://railway.com/deploy/9router-3 --- # Deploy 9Router 9Router AI gateway: one OpenAI-compatible API for 60+ LLM providers **[Deploy 9Router on Railway](https://railway.com/template/9router-3)** Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/9router-3/manifest.json - **Creator:** Will - **Category:** AI/ML ## Template content ### 9Router https://bogusz.co/external/9router-icon.png - **Image:** decolua/9router:0.5.75@sha256:7c893bc2c27ecea2ae337abd5eacfec9e5763091b3a3b7862fc0625b770bb156 - **Health check:** /api/health - **Public domain:** Yes ## Documentation # Deploy and Host 9Router with Railway 9Router is an open-source, self-hosted AI gateway: one OpenAI-compatible endpoint (`/v1`) in front of 60+ LLM providers - Claude, OpenAI, Gemini, DeepSeek, OpenRouter, Ollama, GitHub Copilot, Codex, Antigravity and more - with combos (fallback and round-robin across providers), quota tracking, token savers and a web dashboard. This template deploys the official 9Router image, digest-pinned, with a persistent volume and a deployment healthcheck. **The dashboard password is generated for you**: after deploying, open the `9Router` service, go to **Variables** and copy `INITIAL_PASSWORD`. The login page still says "Default password is 123456" until you save a password of your own - ignore that line and use `INITIAL_PASSWORD`. **Plan requirements.** 9Router is light: measured on this image, the container idles at about **85 MB** and peaked at about **95 MB** after five idle minutes and a burst of 60 concurrent `/v1/models` requests. It runs on every Railway plan, including Free (0.5 GB) and Trial (1 GB). Budget for the service plus whatever your LLM providers charge; the gateway itself adds cents per month. ## About Hosting 9Router 9Router runs as a single Node.js service with a SQLite database under `/app/data`, which this template mounts on a Railway volume so providers, API keys, combos and usage history survive redeploys and restarts. Every secret (`INITIAL_PASSWORD`, `JWT_SECRET`, `API_KEY_SECRET`, `MACHINE_ID_SALT`) is generated at deploy time; nothing has to be typed into the deploy form. The image is pinned to a specific version and digest (`decolua/9router:0.5.75`), so a redeploy never picks up an untested upstream build. Railway probes `/api/health` before switching traffic, and the service restarts on failure. ## Common Use Cases - **One endpoint for every coding agent.** Point Claude Code, Cursor, Cline, OpenCode, Codex or any OpenAI-compatible client at `https://your-service.up.railway.app/v1` with one 9Router API key. - **Fallback and round-robin across providers.** Build a combo that tries your preferred provider first and fails over when it errors or hits a quota. - **Use subscriptions you already pay for.** Connect Claude Code, ChatGPT/Codex, GitHub Copilot or Gemini CLI accounts through OAuth and route requests through them. - **A private gateway for your own apps.** Reach it over Railway's private network as `http://9router.railway.internal:20128` without exposing it. ## Dependencies for 9Router Hosting None. One service, one volume. Provider credentials are added in the dashboard after deploying. ### Deployment Dependencies - [9Router GitHub repository](https://github.com/decolua/9router) - [9Router Docker documentation](https://github.com/decolua/9router/blob/master/DOCKER.md) - [9Router environment variables](https://github.com/decolua/9router#environment-variables) - [Docker Hub image](https://hub.docker.com/r/decolua/9router) ### Implementation Details **Deploying.** Click deploy; there are no inputs. The deployment goes green once `/api/health` answers (about 10-40 seconds), then open the service URL. **Logging in.** Open the service URL, which redirects to `/login`. The password is `INITIAL_PASSWORD` from the `9Router` service's **Variables** tab. Disregard the "Default password is 123456" hint - it only checks whether a password has been saved in the dashboard, not whether one was set through the environment. Then set your own password under **Profile** in the dashboard: leave the *current password* field empty the first time (the upstream form accepts only an empty value or `123456` there before a password has been saved). Once saved, the dashboard password lives in the database and `INITIAL_PASSWORD` is no longer used. **API keys.** `/v1/*` always requires a 9Router API key on a remote deployment - requests without one get `401 API key required for remote API access`. Create keys in the dashboard under **Endpoint** (or `POST /api/keys`) and pass them as `Authorization: Bearer sk-...`. **Connecting providers.** Dashboard → **Providers**. API-key providers (OpenRouter, OpenAI, Anthropic, DeepSeek, ...) take a key and are ready immediately. OAuth providers (Claude Code, Codex, Antigravity, GitHub Copilot, Gemini CLI, ...) open the vendor's consent page. Known upstream issue on any reverse-proxied deployment ([decolua/9router#4054](https://github.com/decolua/9router/issues/4054)): some OAuth flows send the browser back to `http://localhost:443/callback?...` instead of your Railway domain. The code is still valid - replace `http://localhost:443` in the address bar with `https://your-service.up.railway.app` and press Enter to finish the connection. **Coding agents.** Claude Code: ```bash export ANTHROPIC_BASE_URL="https://your-service.up.railway.app" export ANTHROPIC_AUTH_TOKEN="sk-your-9router-api-key" ``` Any OpenAI-compatible client uses `https://your-service.up.railway.app/v1` with the same key. `GET /v1/models` lists the models available from your connected providers. **Persistence and redeploys.** `/app/data` is a Railway volume: `db/data.sqlite`, automatic backups under `db/backups/`, certificates and the model catalog all live there. Verified on this template: API keys and provider connections created before a redeploy are present after it, and a dashboard session stays valid across redeploys because `JWT_SECRET` is stable. Because a volume can only be attached to one container at a time, a redeploy has a few seconds of downtime (measured 1 failed probe out of 150 at one-second intervals); in-flight streaming responses are cut at that moment. **Optional variables.** `CLOUD_URL` is only used if you enable Cloud Sync. `ENABLE_REQUEST_LOGS=true` writes per-request logs to `/app/logs`, which is not on the volume and grows without bound - leave it off unless you are debugging. `HEADROOM_URL` and `SEARXNG_URL` point at optional sidecar services (a Headroom token-saver or a SearXNG instance) if you add them to the project; use their private `*.railway.internal` addresses. **Upgrades.** The image is pinned by digest. To move to a newer 9Router, change the service's source image to the new tag (check the [changelog](https://github.com/decolua/9router/blob/master/CHANGELOG.md) and open issues first - upstream has shipped versions where the dashboard broke, and `latest` has moved eight times in six weeks), then redeploy. Data on the volume is preserved; take a copy of `db/data.sqlite` first if you want a rollback point. **Security notes.** The dashboard requires a login and the LLM API requires a key on every request from outside the container. The session cookie is marked Secure (`AUTH_COOKIE_SECURE=true`). `API_KEY_SECRET` and `MACHINE_ID_SALT` are generated so keys and machine identity cannot be derived from upstream's public defaults. Railway's platform terms apply to how you use the gateway; run it for your own accounts and keys. ### Why Deploy 9Router 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 9Router 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. ## Similar templates - [Chat Chat](https://railway.com/deploy/-WWW5r) — Chat Chat, your own unified chat and search to AI platform. - [stella](https://railway.com/deploy/stella) — Self-host stella with web, API, Postgres, Redis, and object storage. - [Hermes Agent | OpenClaw Alternative with Dashboard](https://railway.com/deploy/hermes-agent-or-openclaw-alternative-wit) — Self-Hosted Hermes AI Agent for Telegram, Discord & Slack Open this page in a browser: https://railway.com/deploy/9router-3