Back to Overview
Hostinger VPS Setup

QUICK START GUIDE

Every step to get OpenClaw running on Hostinger VPS — from provisioning to first AI response. Check off each step as you go.

0%
0 / 62 steps
Phase 1

Hostinger VPS — Provision & Deploy

1

Go to hostinger.com/vps-hosting and click Get Started.

2

Select a VPS plan — minimum KVM 1 (1 vCPU, 4 GB RAM) is sufficient for OpenClaw.

3

Under OS / Application, choose Application → OpenClaw from the dropdown.

4

Set your server location to the region closest to you or your clients.

5

Complete checkout and wait for the VPS to provision (typically 2–5 minutes).

6

Open hPanel → VPS → your new server → Overview.

7

Note your VPS IPv4 address — you will need it to access the OpenClaw dashboard.

Phase 2

Environment Variables — API Keys

8

In hPanel → VPS → your server → click Manage → Environment Variables (or Docker Manager → your OpenClaw container → Edit).

9

Locate the OPENCLAW_GATEWAY_TOKEN variable — copy and save this value somewhere safe (this is your dashboard password).

!

Never share your OPENCLAW_GATEWAY_TOKEN — it grants full control of your OpenClaw instance.

11

Go to platform.openai.com/api-keys and click Create new secret key.

12

Name the key 'OpenClaw VPS', set Permissions to All (or restrict to Model capabilities: Write only).

13

Copy the key immediately — it will not be shown again.

14

Back in hPanel, add a new environment variable: OPENAI_API_KEY = your key from step 13.

Optional: Add ANTHROPIC_API_KEY (from console.anthropic.com) and/or GEMINI_API_KEY (from aistudio.google.com) for additional model options.

16

Click Save / Apply — the container will restart automatically.

Phase 3

Access the OpenClaw Dashboard

17

In hPanel → Docker Manager, find your OpenClaw container and note the assigned port (default is 18789).

18

Open a browser and navigate to http://<your-vps-ip>:18789 — replace <your-vps-ip> with the IPv4 from step 7.

http://123.456.78.90:18789
19

Enter your OPENCLAW_GATEWAY_TOKEN from step 9 when prompted for the password.

20

Confirm the Gateway status shows Running (green indicator) on the dashboard home screen.

Phase 4

Connect WhatsApp

21

In the OpenClaw dashboard, click Channels in the left sidebar.

22

Click Show QR next to the WhatsApp channel.

23

On your phone, open WhatsApp → Settings → Linked Devices → Link a Device.

24

Scan the QR code displayed on screen — the channel status will change to Connected.

If you see a status=515 error: go to Settings → Config → Update in the dashboard, then return to Channels and try again.

26

Send a test message to your WhatsApp number and confirm the AI responds.

Phase 5

Connect Telegram (Optional)

27

Open Telegram and search for @BotFather.

28

Send /newbot to BotFather and follow the prompts to name your bot.

/newbot
29

Copy the bot token BotFather provides (format: 123456789:ABCdefGHIjklMNOpqrSTUvwxYZ).

30

In the OpenClaw dashboard → Channels → Telegram, paste your bot token and click Connect.

31

Message your bot in Telegram and confirm the AI responds.

Phase 6

Select Your AI Model

32

In any connected channel (WhatsApp or Telegram), send the command:

/model list
33

Review the available models — your configured API keys determine which are available.

34

Set your default model by sending:

/model openai/gpt-4.1-mini

Recommended starting model: gpt-4.1-mini (fast, cost-efficient). Upgrade to gpt-4o for more complex tasks.

36

Send a test message to confirm the model is active and responding correctly.

Phase 7

Install Skills from ClawHub

37

Visit clawhub.ai and browse available skills.

38

Find the API Gateway skill — this is required for all external API integrations (email, search, etc.).

39

Copy the install command from the skill's page on ClawHub.

40

In your connected chat (WhatsApp or Telegram), paste and send the install command.

41

Confirm the skill installed successfully — the bot will reply with a confirmation message.

For cold outreach: also install the Brave Search skill. Get your free Brave Search API key at brave.com/search/api.

Phase 8

OpenAI API Key — Permissions

43

Go to platform.openai.com → Projects and create a new project named 'OpenClaw'.

44

Inside the OpenClaw project, go to API Keys → Create new secret key.

45

Set key name to 'OpenClaw VPS'.

46

Set Permissions to Restricted → enable Model capabilities: Write only.

47

Leave Fine-tuning, Assistants, Files, Billing, and Organization all set to None.

48

Click Create secret key and copy it immediately.

49

Go to Billing → Limits and set a monthly spend limit (e.g. $20) as a safeguard.

50

Update OPENAI_API_KEY in your Hostinger environment variables (hPanel → Docker Manager → Edit container) with this new scoped key.

Phase 9

Settings Instructions — Customize Your Agent

OpenClaw uses Markdown files in the workspace (~/.openclaw/workspace/) as the agent's instructions. Edit these files to control behavior, persona, and memory.

57

In your connected chat, send the command to open the workspace file editor:

/edit AGENTS.md
58

In AGENTS.md — write your operating rules: priorities, response style, tasks the agent should always/never do.

59

Send the command to edit the persona file:

/edit SOUL.md
60

In SOUL.md — define the agent's tone, personality, and communication boundaries (e.g. 'Always be concise. Never use emoji.').

61

Send the command to set user context:

/edit USER.md
62

In USER.md — describe who you are, your business, preferred name, and any personal context the agent should always know.

63

Send the command to set the agent's identity:

/edit IDENTITY.md
64

In IDENTITY.md — set the agent's name, emoji, and vibe (e.g. 'Name: Aria | Emoji: 🤖 | Vibe: Professional, direct').

Keep all workspace files concise — every file is injected into the context window on every turn and consumes tokens. Aim for under 500 words per file.

Phase 10

Heartbeat — Proactive Mode

Heartbeat runs a background agent turn on a timer — no message from you needed. The agent checks HEARTBEAT.md and proactively sends you alerts or updates.

!

Heartbeat is ON by default (every 30 minutes). Disable it first until you have tested your setup to avoid unexpected messages and token usage.

68

Disable heartbeat immediately by sending:

/config set agents.defaults.heartbeat.every 0m
69

Edit your HEARTBEAT.md to define what the agent should check on each run:

/edit HEARTBEAT.md
70

In HEARTBEAT.md — write a short checklist (e.g. 'Check if any leads replied. Summarize any calendar events due today. If nothing, reply HEARTBEAT_OK.').

If the agent has nothing to report it replies HEARTBEAT_OK — this is silently suppressed and you will NOT receive a notification.

72

Re-enable heartbeat with your preferred interval once HEARTBEAT.md is configured:

/config set agents.defaults.heartbeat.every 30m
73

Restrict heartbeat to business hours only to avoid off-hours pings — add to openclaw.json:

heartbeat: { every: "30m", activeHours: { start: "09:00", end: "22:00" } }

Cost tip: set heartbeat to use a cheaper model (e.g. gemini-2.5-flash) to reduce API costs — heartbeat runs do not need a powerful model.

Phase 11

Ongoing Maintenance

51

To update OpenClaw: hPanel → Docker Manager → your container → Update Image.

52

To view logs: hPanel → Docker Manager → your container → Logs.

53

To restart the gateway: hPanel → Docker Manager → your container → Restart.

54

To back up your config: download the environment variables from hPanel and store them securely.

OpenClaw auto-restarts on VPS reboot — no manual intervention needed after a server restart.