Deploy and Host NextChat on Railway

Pinned NextChat (ChatGPT-Next-Web) on Railway: official Docker image, generated access code, optional OpenAI-compatible backend.

This listing deploys yidadaa/chatgpt-next-web:v2.16.1 (manifest sha256:eaaa469ddeeb5fa58fb35f8767e9e096a2f1c8468c54b6703323624bf3071c5a) — the current image published by ChatGPTNextWeb/NextChat. It does not Nix-build the redirected Yidadaa/ChatGPT-Next-Web GitHub URL.

About Hosting NextChat

NextChat is a self-hosted ChatGPT-style UI. It proxies requests to OpenAI or any OpenAI-compatible backend and stores chat history in the browser, so no database or volume is required. Hosting your own instance lets you generate a shared CODE access password, optionally embed OPENAI_API_KEY, and point BASE_URL at LiteLLM, LocalAI, vLLM, OpenRouter, or a custom proxy.

One-click deploy:

Click Deploy on Railway.
Railway generates CODE. Copy it from the service variables after deploy.
Optionally paste OPENAI_API_KEY, or leave it empty and enter a key in the UI.
Optionally set BASE_URL for an OpenAI-compatible API.
Open the public HTTPS domain and enter CODE when prompted.

Common Use Cases

Shared team UI — one generated access code, optional server-side API key so teammates do not each manage credentials.
OpenAI-compatible frontend — set BASE_URL at a self-hosted or third-party /v1 endpoint (LiteLLM, LocalAI, vLLM, OpenRouter).
Locked-down public domain — unlike clones that ship no CODE, this template generates an access password so the HTTPS URL is not an open proxy.

Dependencies for NextChat Hosting

None required to boot the UI. An API key for at least one provider is needed to chat (OpenAI via OPENAI_API_KEY, another vendor in Settings, or BASE_URL plus that backend's key).
No volume. Chats live in the browser (export/import or WebDAV if you want backups).
Port 3000 with a Railway public HTTP domain.

Deployment Dependencies

Image: yidadaa/chatgpt-next-web:v2.16.1 (ChatGPTNextWeb/NextChat, Docker Hub yidadaa/chatgpt-next-web)
Upstream environment variables

Variable	Required	Default	Notes
`CODE`	generated	`${{secret()}}`	Access password. Comma-separate multiple codes.
`OPENAI_API_KEY`	optional	empty	Server-side OpenAI key. Leave empty to paste a key in Settings.
`BASE_URL`	optional	empty (`https://api.openai.com`)	OpenAI-compatible API base URL.
`HOSTNAME`	set	`0.0.0.0`	Bind address inside the container.
`PORT`	set	`3000`	HTTP listen port.

Optional provider keys you can add later: GOOGLE_API_KEY, ANTHROPIC_API_KEY, AZURE_URL + AZURE_API_KEY, DEEPSEEK_API_KEY.

Login: open the Railway HTTP domain. NextChat asks for Access Code — that is the generated CODE variable.

Healthcheck: the image has no dedicated /health path. Railway GET / healthchecks fail on Next.js host checks (healthcheck.railway.app). Verify with curl of the public domain: GET / returns HTTP 200 and ``.

Why Deploy NextChat on Railway?

The search-winning NextChat card still Nix-builds Yidadaa/ChatGPT-Next-Web (GitHub redirect), ships no access code, and has no healthcheck (health ~50). This template pins the current ChatGPTNextWeb/NextChat Docker tag (not :latest, not a Nixpacks GitHub clone), generates CODE, leaves OPENAI_API_KEY empty/optional, documents BASE_URL, and restarts on failure.

Railway hosts the container, public HTTPS domain, and generated secret so you do not have to run Docker yourself.