--- title: "Deploy OpenClaw — Private AI Agent" description: "Self-host OpenClaw: private AI agent for ~$1/month. No ChatGPT subscription" category: "AI/ML" url: https://railway.com/deploy/openclaw-private-agent --- # Deploy OpenClaw — Private AI Agent Self-host OpenClaw: private AI agent for ~$1/month. No ChatGPT subscription **[Deploy OpenClaw — Private AI Agent on Railway](https://railway.com/template/openclaw-private-agent)** - **Creator:** SilverBanana - **Category:** AI/ML - **Total deploys:** 9 ## Template content ### Openclaw https://d1dm0o5huagnd6.cloudfront.net/static/openclaw.jpeg - **Source:** sahilrupani/openclaw-railway-template - **Public domain:** Yes ## Documentation # Deploy and Host OpenClaw on Railway ![OpenClaw AI Agent](https://opengraph.githubassets.com/027231e34679f13d043884e2d69bd69e052e500e3bf7b5b03c72101eda21b724/openclaw/openclaw) OpenClaw (formerly Clawdbot and Moltbot) is the open-source AI agent with 328k+ GitHub stars that goes beyond chat — it browses the web, manages files, runs commands, handles email, and works autonomously on your behalf, accessible from WhatsApp, Telegram, Discord, Slack, iMessage, and 20+ other messaging platforms. **Self-host a personal AI agent for ~$1–3/month** — compared to $20/month for ChatGPT Plus, $20/month for Claude Pro, or Manus AI which is cloud-only with no self-host option. Full data ownership. No usage caps. No subscription. --- ## What This Template Deploys | Service | Purpose | |---|---| | **OpenClaw Gateway** | Core agent runtime — LLM routing, channel connections, autonomous task execution, and Control UI at `/openclaw` | | **Persistent Volume (5 GB)** | Railway volume at `/data` — config, credentials, memory, and workspace files survive every redeploy | No database. No Redis. No inter-service networking. State lives entirely on the persistent volume — the simplest production OpenClaw stack on Railway. --- ## About Hosting OpenClaw Running OpenClaw in production means keeping a persistent Node.js gateway alive 24/7, managing secure credentials for your LLM provider and messaging platforms, handling HTTPS termination, and ensuring config and memory survive redeploys. Without a managed host, you're looking at VPS setup, Nginx config, SSL certificates, process managers, and manual Docker maintenance. Railway handles all of it. This template mounts a **5 GB persistent volume** at `/data`, provisions the gateway with automatic HTTPS, and injects secrets as environment variables. OpenClaw's memory, channel auth, and workspace survive every redeploy without manual intervention. The browser-based `/setup` wizard configures everything after deploy — no SSH, no terminal, no config files. Typical cost: **~$1–3/month** on Railway's Hobby plan. ChatGPT Plus costs $20/month. Claude Pro costs $20/month. Manus AI is cloud-only with no self-host option. OpenClaw on Railway gives you an autonomous AI agent at a fraction of the cost with full data ownership. --- ## Deploy in Under 5 Minutes 1. Click **Deploy on Railway** and let the initial build complete (~3–5 minutes) 2. Open your Railway URL and complete the `/setup` wizard — pick your AI provider and paste your API key 3. Optionally connect a messaging channel — Telegram bot token, Discord bot token, or Slack app 4. Start chatting from your browser or messaging app No SSH. No command line. No server administration. ![OpenClaw setup wizard — configure your AI provider and messaging channels](https://res.cloudinary.com/asset-cloudinary/image/upload/v1774088297/openclaw_setup_page_rxowv8.png) --- ## Common Use Cases - **Self-hosted alternative to ChatGPT Plus and Claude Pro** — your conversations stay on your Railway instance, not OpenAI or Anthropic servers; swap providers freely without losing memory or configuration - **Self-hosted alternative to Manus AI** — get autonomous agent capability on infrastructure you control without a cloud-only dependency or third-party data exposure - **Personal AI assistant via WhatsApp or Telegram** — message your agent to browse the web, summarize documents, draft emails, or run scheduled tasks from any messaging app you use - **Autonomous task scheduling** — configure the heartbeat daemon to run recurring tasks (daily briefings, price monitoring, data pipelines) on a schedule without manual prompting - **Multi-platform bot** — connect one agent to Telegram, Discord, and Slack simultaneously with shared memory and session context across all channels - **Private LLM switching** — test Anthropic, OpenAI, Gemini, and OpenRouter behind a consistent interface without reconfiguring your setup each time you switch providers --- ## Configuration | Variable | Required | Description | |---|---|---| | `ANTHROPIC_API_KEY` | Recommended | API key for Anthropic Claude — best results for autonomous tasks and long-context reasoning | | `OPENAI_API_KEY` | Alternative | API key for OpenAI GPT models | | `GEMINI_API_KEY` | Alternative | API key for Google Gemini | | `OPENROUTER_API_KEY` | Alternative | Single key for multiple providers — useful for fallback routing | | `SETUP_PASSWORD` | Recommended | Protects the `/setup` wizard — set before sharing your Railway URL | | `OPENCLAW_GATEWAY_TOKEN` | Auto-generated | Secure UI-to-gateway token — generated at deploy time | | `TELEGRAM_BOT_TOKEN` | Optional | From BotFather — enables Telegram channel | | `DISCORD_BOT_TOKEN` | Optional | From Discord Developer Portal — enables Discord channel | | `SLACK_BOT_TOKEN` | Optional | From Slack API — enables Slack channel | | `OPENCLAW_STATE_DIR` | Pre-set | Defaults to `/data/.openclaw` — do not change unless migrating | | `OPENCLAW_VERSION` | Optional | Pin to a specific release e.g. `v2026.2.16` — omit for latest | > `PORT` is injected automatically by Railway. Do not set it manually. You only need **one API key** to get started. Everything else is configurable through the `/setup` wizard after deploy. --- ## OpenClaw vs. Alternatives | | OpenClaw (self-hosted) | ChatGPT Plus | Claude Pro | Manus AI | |---|---|---|---|---| | **Monthly cost** | ~$1–3 (Railway) | $20/month | $20/month | Cloud-only | | **Self-hostable** | ✅ Yes | ❌ No | ❌ No | ❌ No | | **Data ownership** | ✅ Your instance | ❌ OpenAI servers | ❌ Anthropic servers | ❌ Cloud servers | | **Usage caps** | ✅ None | ❌ Rate limited | ❌ Rate limited | ❌ Yes | | **LLM provider choice** | ✅ Any provider | ❌ OpenAI only | ❌ Anthropic only | ❌ Fixed | | **Messaging platforms** | ✅ 20+ platforms | ❌ Web only | ❌ Web only | ❌ Web only | | **Autonomous scheduling** | ✅ Heartbeat daemon | ❌ No | ❌ No | ✅ Limited | | **Open source** | ✅ Yes | ❌ No | ❌ No | ❌ No | --- ## Dependencies for OpenClaw Hosting - API key from at least one LLM provider — Anthropic Claude recommended - Railway account (Hobby plan — ~$5/month base, template itself ~$1–3/month) - Optional: bot tokens for Telegram, Discord, or Slack ### Deployment Dependencies - [Official OpenClaw GitHub Repository](https://github.com/openclaw-ai/openclaw) - [OpenClaw Documentation](https://docs.openclaw.ai/) - [Railway Volumes Documentation](https://docs.railway.com/volumes) - [Anthropic API Keys](https://console.anthropic.com/) - [Telegram BotFather](https://t.me/BotFather) - [Discord Developer Portal](https://discord.com/developers/applications) ### Implementation Details This template deploys OpenClaw's gateway as a persistent container with a **5 GB Railway volume** at `/data` — larger than most competing OpenClaw templates on Railway. All agent state lives on the volume: LLM provider config, channel auth tokens, persistent memory, and workspace files. The container is fully stateless — redeploy, version update, or restart without losing anything. The gateway exposes a Control UI at `/openclaw` and a setup wizard at `/setup`, both protected by token-based authentication. No database or Redis instance required — all state is filesystem-based on the persistent volume, keeping the stack simple and the cost low. --- ## Frequently Asked Questions **Is OpenClaw the same as Clawdbot or Moltbot?** Yes. OpenClaw was previously known as Clawdbot and before that as Moltbot. The project was rebranded to OpenClaw — same team, same codebase, continued development. Existing Clawdbot or Moltbot users can migrate data by moving the `/data` volume contents to the new instance. **How much does OpenClaw cost on Railway vs ChatGPT or Claude?** The Railway container costs ~$1–3/month on the Hobby plan. ChatGPT Plus is $20/month with rate limits. Claude Pro is $20/month with usage caps. LLM API costs are separate and depend on usage — Claude Haiku or GPT-4o-mini keep inference costs low for personal use. **What makes this template different from other OpenClaw Railway templates?** This template ships with a **5 GB persistent volume** — larger than the standard 1–2 GB most competing templates configure. The `/setup` wizard handles all channel configuration, provider selection, and credential management in the browser. No SSH, no config files, no token generation required. **Is my data private?** Yes. OpenClaw is fully self-hosted. Conversations stay on your Railway instance. API keys are stored as encrypted Railway environment variables. No data is sent to third-party analytics services. **Do I lose config and memory when Railway redeploys?** No. All state is stored on the persistent volume at `/data`, not inside the container. Redeploys, version updates, and restarts do not affect your data. **How do I update OpenClaw to a newer version?** Set `OPENCLAW_VERSION` to the target release (e.g. `v2026.2.16`) in Railway Variables and redeploy. Omit it or set to `latest` to always pull the newest release on each deploy. --- ## Why Deploy and Host OpenClaw 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. At ~$1–3/month, OpenClaw on Railway is the cheapest persistent AI agent stack available — own your data, control your LLM provider, and run 24/7 autonomous tasks without a subscription or usage cap. ## 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/openclaw-private-agent