OpenClaw Deploy

OpenClaw Deploy lets you spin up a personal AI Telegram bot in under 60 seconds — powered by Claude, GPT-4, Gemini, or a locally-hosted Ollama model. Each bot runs in an isolated Docker container on dedicated Hetzner infrastructure. You bring your own API key; we handle the hosting.

What you get

• A live Telegram bot at t.me/YourBotName
• Full OpenClaw agent capabilities — web search, memory, code execution, tools
• Isolated container — your bot doesn't share memory or context with other users
• A dashboard to monitor status, view logs, stop or redeploy
• Stripe-managed subscription — cancel any time

Requirements

• A Google account (used for authentication)
• A Telegram bot token from @BotFather
• An API key for your chosen AI provider (not required for Ollama)
• An active Basic or Pro subscription

Quick start

From zero to a live bot in four steps.

Supported models

Pick any provider at deploy time. You can redeploy with a different model at any time.

Available Ollama models (on-VPS)

How it works

Architecture

Each user gets an isolated Docker container running the openclaw-headless:latest image — a full OpenClaw agent configured to listen on Telegram.

  User browser
      │
      ▼  POST /api/deploy  (HTTPS, Nginx → Express :3200)
  openclaw-deploy API  ──► Docker run ──► Container openclaw-user-{id}
      │                                        │
      ▼  SQLite DB                             ▼
  deployments table                    Telegram Bot API
  (no API key stored)                  ← polls for messages

Container lifecycle

• Deploy: Any existing container for your user is stopped first, then a fresh one is launched with your credentials injected as environment variables.
• Running: Container polls Telegram for messages and responds using your model/key.
• Stop: Container is stopped but the deployment record is retained. You can redeploy at any time.
• Redeploy: Same flow as deploy — enter credentials again (API key is never cached).

Infrastructure

• VPS: Hetzner nbg1, 24 GB RAM, Ubuntu 24 — CPU-only
• Max containers: 30 simultaneous running bots per VPS
• Networking: Each container connects outbound-only (to Telegram and your AI provider). Port 18789 (OpenClaw gateway) is never exposed externally.
• SSL: TLS via Let's Encrypt, auto-renewing. All API traffic is HTTPS.

Security & privacy

What we store

Authentication

We use Google Identity Services (OAuth 2.0). No passwords are stored. Sessions are JWT cookies (httpOnly, SameSite=lax, 30-day expiry). Sign out clears the cookie immediately.

Container isolation

Each container runs with only the environment variables it needs. Docker networking ensures containers cannot communicate with each other. The internal Ollama service is bound to 172.17.0.1 (Docker bridge) and is not reachable from the public internet.

User dashboard

Once you sign in and have an active deployment, the deploy form is replaced by your dashboard.

Status indicator

• ● Running — bot is live and responding to messages
• ● Starting — container is booting (usually <30s)
• ● Stopped — container is not running; bot is offline

Actions

Subscription badge

An active Basic or Pro plan shows a ✦ pro badge next to your status. If no badge appears, your subscription may be inactive — check the pricing section to renew.

API reference

The backend runs at https://openclaw.aiindigo.com. All endpoints require a valid JWT cookie (set via POST /api/auth/google) unless marked public.

Authentication

{
  "success": true,
  "user": { "id": 1, "email": "you@gmail.com", "name": "Your Name", "picture": "https://..." }
}

Deployment

{
  "success": true,
  "containerId": "a1b2c3d4...",
  "botName": "@YourBot",
  "botUrl": "https://t.me/YourBot",
  "status": "starting"
}

Status & monitoring

{
  "id": 1,
  "status": "running",  // "running" | "stopped" | "starting" | "not_deployed"
  "model_provider": "anthropic",
  "model_name": "claude-sonnet-4-5",
  "bot_name": "@YourBot",
  "container_name": "openclaw-user-1",
  "created_at": "2026-02-22T14:00:00.000Z",
  "container": { "running": true, "status": "running" }
}

{ "success": true, "message": "Container stopped" }

Subscription

{
  "subscription": {
    "plan": "pro",
    "status": "active",
    "current_period_end": 1774000000
  }
}

{ "url": "https://checkout.stripe.com/..." }

Public config

{ "googleClientId": "123....apps.googleusercontent.com" }

Pricing

You pay for infrastructure hosting. AI model costs are yours (via your own API key).

FAQ