--- title: "Deploy LLM Wiki" description: "Karpathy's LLM wiki, self-hosted: Claude builds your wiki over MCP" category: "AI/ML" url: https://railway.com/deploy/llm-wiki --- # Deploy LLM Wiki Karpathy's LLM wiki, self-hosted: Claude builds your wiki over MCP **[Deploy LLM Wiki on Railway](https://railway.com/template/llm-wiki)** Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/llm-wiki/manifest.json - **Creator:** Youssef Siam - **Category:** AI/ML ## Template content ### api - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-api:1.0.1 - **Health check:** /health - **Public domain:** Yes ### storage - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-storage:1.0.1 - **Public domain:** Yes ### kong - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-kong:1.0.1 - **Public domain:** Yes ### web - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-web:1.0.1 - **Health check:** /login - **Public domain:** Yes ### db - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-db:1.0.1 ### converter - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-converter:1.0.1 - **Health check:** /health ### auth - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-auth:1.0.1 ### mcp - **Image:** ghcr.io/youssefsiam38/llmwiki-railway-mcp:1.0.1 - **Health check:** /health - **Public domain:** Yes ## Documentation # Deploy and Host LLM Wiki on Railway LLM Wiki is an open-source take on Andrej Karpathy's LLM wiki idea: a personal Wikipedia that an AI builds and maintains for you. Upload papers, notes, PDFs, Word and PowerPoint files, connect Claude (or Codex, or any MCP client), and let it compile interlinked wiki pages with citations back to your sources, then keep them current as you add more. This is a community-maintained template; it is not affiliated with the LLM Wiki project. ## About Hosting LLM Wiki LLM Wiki's hosted mode is a Next.js web app, a FastAPI API, an MCP server and a document converter (LibreOffice and opendataloader), on Supabase Auth and Postgres with S3 storage. Claude and other MCP clients sign in through Supabase Auth's OAuth 2.1 server. Upstream runs it on Supabase Cloud and AWS and documents only its local desktop mode. This template runs the whole hosted stack on Railway, in one project, eight services with every secret generated: a self-hosted Supabase (Postgres, Auth with the OAuth server, and its gateway), RustFS object storage, the converter, and LLM Wiki's API, MCP server and web app, built from a pinned upstream commit. No Supabase account and no AWS. The first start does the setup for you: it applies LLM Wiki's database migrations, creates the storage bucket with a CORS policy for your web app, creates your owner account from the e-mail you enter, and only then starts the API. Signup is closed by default and enforced in the database: only the owner and the addresses or domains you list get accounts. ## Why Deploy LLM Wiki 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 LLM Wiki 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. Concretely, this template keeps the database, Supabase Auth and the converter on Railway's private network, serves the web app, API, MCP endpoint, sign-in gateway and signed file links over HTTPS, and attaches volumes to the database and the document store. Your MCP server has a stable HTTPS URL that Claude's custom connectors can reach. ## Common Use Cases - A research wiki that Claude keeps up to date from the papers and articles you upload. - A team's institutional memory: shared sources, compiled pages, one account per colleague. - Course notes and study guides built from lecture PDFs and slides. - A private context layer for Claude, Codex or Cursor over MCP, on infrastructure you control. ## Dependencies for LLM Wiki Hosting - An MCP client such as Claude (claude.ai custom connectors, Claude Desktop or Claude Code), Codex or Cursor. The template itself needs no API key. - Optional: a Mistral API key for higher-quality PDF OCR. - Optional: SMTP for password-reset e-mails, and a Google OAuth client for Google sign-in. ### Deployment Dependencies - LLM Wiki: https://github.com/lucasastorian/llmwiki (Apache-2.0) - Supabase self-hosting stack: https://github.com/supabase/supabase/tree/master/docker (Apache-2.0) - Supabase Auth: https://github.com/supabase/auth (MIT) - RustFS: https://github.com/rustfs/rustfs (Apache-2.0) - Template repository, images and tests: https://github.com/youssefsiam38/llmwiki-railway ### Implementation Details The API, MCP server and web app are built from a pinned upstream commit with upstream's hash-locked dependencies and its pip-audit gate; the web build raises the dependency pins that carry published advisories, including a critical Next.js one. Supabase Auth signs sessions with an ES256 key derived from a generated seed, which LLM Wiki's API and MCP server require and Railway cannot generate directly. The converter's download allowlist is narrowed to the template's own storage bucket, and the API stores files over the private network while signing browser links with the public storage domain. Every service refuses to start on missing, short or published example secrets. The bundle is tested as a whole in CI and on a live deployment of this template: sign-in, the signup gate, a PDF extracted by the converter, signed file links, account isolation, and a complete MCP session as Claude runs it (client registration, consent, token exchange, tool calls that read the PDF and write and search a wiki page, token refresh and revocation), on fresh and reused volumes. The deploy form asks for one value, `OWNER_EMAIL`. After deploying, copy `OWNER_PASSWORD` from the api service's variables, sign in on the web app's domain, and add the MCP URL from the settings page to Claude as a custom connector. ## 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/llm-wiki