OpenClaw on Hostinger: Complete 2026 Managed Setup Guide

Get a working OpenClaw agent running on Hostinger Managed OpenClaw in under 10 minutes — no Linux, no SSH, no terminal. This guide walks through every step, covers honest cost math (hosting + LLM API combined), shows how to connect Telegram and Discord, and explains the security caveats Hostinger's sales page doesn't mention.

Affiliate disclosure: we earn a small commission when you sign up via our Hostinger link, and you get 20% off the listed price — the discount is applied automatically when you click through (no coupon code to copy). Full disclosure.

What OpenClaw is (and what it isn't)

OpenClaw is an open-source AI agent framework — a runtime that lets a large language model actually do things on your behalf rather than just chat. It was released in November 2025 by Austrian developer Peter Steinberger (originally as "Clawdbot") and rebranded after trademark complaints. Today the project has ~247,000 GitHub stars and is sponsored by OpenAI, GitHub, and NVIDIA, with an MIT license.

OpenClaw is not a chatbot you talk to on a website. You access it through messaging apps you already use — Telegram, Discord, Slack, Signal, iMessage, or WhatsApp — and it runs continuously on a server, executing tasks autonomously. Typical uses:

• "Triage my Gmail, summarise the important threads, and reply to the routine ones."
• "Watch my Telegram for messages from $person and reply if it's urgent."
• "Every morning at 8 AM, send me a Discord DM with the weather, my calendar, and overnight Bitcoin price."
• "Process this folder of PDFs and put a summary in my Notion."

It's BYOK (bring your own key) for the actual language model — you supply credentials to Anthropic, OpenAI, Google, or DeepSeek. The OpenClaw framework itself is free; you only pay for hosting and for whatever tokens your chosen LLM consumes.

Why Managed OpenClaw is the right pick for most people

You technically have other options — you could rent a Linux VPS, install OpenClaw via the official curl installer, configure systemd, set up Caddy for SSL, install fail2ban, and maintain the OS yourself. People do that, and it works. But unless you genuinely enjoy SSH and OS administration, the managed product is a better fit for almost every reader of this guide.

Here's the honest comparison:

Managed OpenClaw saves you a few dollars worth of complexity and several hours of setup, in exchange for a couple of dollars more per month. For anyone who isn't already running a fleet of self-managed VPSes, it's the obviously better trade. The rest of this guide covers Managed OpenClaw exclusively.

If you want a precise dollar comparison for your specific workload (managed hosting + your chosen LLM backend + your usage profile), use the OpenClaw Cost Calculator.

Prerequisites

Before you start, gather:

1. A Hostinger account. Free to create. We use Hostinger because their Managed OpenClaw is one of the few first-party offerings of this product in 2026 and the pricing is competitive. Signing up via our link auto-applies a 20% discount at checkout (and earns us a small commission).
2. An LLM API key. Required for OpenClaw to actually do anything — the framework needs a brain. The most common choices in 2026:
   • Anthropic Claude (recommended for most users) — sign up at console.anthropic.com, add billing, create an API key. Sonnet 4.6 is the sweet spot at $3/$15 per million tokens.
   • OpenAI GPT-5 — sign up at platform.openai.com, add billing, create an API key. GPT-5.4 Mini at $0.25/$1.00 per million tokens is the cheap reliable option.
   • Google Gemini — get a key at ai.google.dev. Gemini 3 Flash is shockingly cheap at $0.07/$0.30 per million.
   • DeepSeek V3.2 — at platform.deepseek.com. $0.28/$1.10 per million, very strong on code tasks.
3. A chat surface for talking to your agent. You don't need this on day one — OpenClaw has a built-in web UI — but most users want their agent on Telegram or Discord. Free to set up either.
4. ~10 minutes of focus. Most of that is waiting for the instance to provision and for your LLM provider to confirm billing.

Step-by-step setup (8 steps, ~10 minutes)

This is the full walkthrough from "I just decided to try OpenClaw" to "my agent just responded to me on Telegram." Hostinger handles install, updates, SSL, and backups — your only responsibilities are the LLM API key, the persona, and the chat surface.

Step 1: Sign up and pick a plan

Head to Hostinger Managed OpenClaw. There are two tiers as of May 2026:

• Starter ($11.99/mo annual) — for one personal agent with light/medium usage. The right pick for almost everyone reading this guide.
• Power ($14.99/mo annual) — multiple agents, heavier workloads, more concurrent sessions.

Pick annual billing. Monthly billing is roughly double the annual rate. You can upgrade tiers later without reinstalling — Hostinger preserves your config and persona.

Step 2: Provision the instance

After checkout, Hostinger spins up your instance automatically. You'll get an email when it's ready, usually within 60-120 seconds. Click Manage in your Hostinger dashboard to reach the OpenClaw panel for the first time.

Step 3: Set your admin password

The panel prompts you to create an admin password on first login. Use a password manager — Bitwarden, 1Password, or your browser's built-in manager. This password protects your agent's web UI and is independent from your Hostinger account password. Anyone who gets this password can read your agent's conversation history and run commands as your agent, so treat it as sensitive.

Step 4: Add your LLM API key

In the panel, navigate to Settings → LLM Provider. Pick your provider from the dropdown (Anthropic, OpenAI, Google, DeepSeek) and paste the API key you created in the prerequisites section. The panel will validate it by making a small test request — you'll know within ~5 seconds if the key is valid and billing is set up correctly.

Choose your model. For first-time setup, we recommend:

• Anthropic → Claude Sonnet 4.6 (best overall balance)
• OpenAI → GPT-5.4 Mini (cheapest reliable option)
• Google → Gemini 3 Flash (cheapest of the four major providers)
• DeepSeek → V3.2 (very strong on code tasks)

You can switch models any time without losing your agent's state or conversation history.

Step 5: Pick your agent persona

The panel ships with several pre-built personas, each a starter system prompt plus a curated set of skills:

• Personal Assistant — calendar, email triage, daily briefings, reminders
• Research Agent — web search, document Q&A, summarisation
• Coding Buddy — code review, git operations, PR summaries
• Inbox Triager — Gmail / Outlook integration, prioritisation, auto-replies

Pick the one that matches your primary use case. You can customise the system prompt and enable/disable individual skills later — the persona is a starting template, not a permanent identity.

Step 6: First test message

The panel includes a built-in chat window. Send something simple to verify everything works end-to-end:

Hello — what skills do you have available?

Your agent should reply within a few seconds with a list of installed skills. If it errors out, jump to the troubleshooting section.

Step 7: Connect a chat surface (optional but recommended)

To use your agent from your phone, connect Telegram or Discord. The Managed panel has a one-click connector for both — you only paste a bot token and authorise your user ID. Detailed step-by-step in the Telegram and Discord sections below.

Step 8: You're done

That's it. Your agent runs 24/7 inside Hostinger's managed environment, restarts itself on crashes, and updates automatically when Hostinger pushes new OpenClaw releases. Your only ongoing responsibilities are:

• Keeping enough credit on your LLM provider account to cover usage
• Reviewing skills before installing third-party ones (some skills request broad permissions)
• Setting an allowed_users list on any messaging connector so random people can't reach your bot

Connecting your LLM API key

OpenClaw is BYOK — you supply your own credentials to the LLM provider. Here's how to get a key from each of the four most-common providers as of 2026.

Anthropic Claude

1. Sign up at console.anthropic.com
2. Go to Plans & Billing and add a payment method. Anthropic requires a $5 minimum credit deposit.
3. Go to API Keys → Create Key. Name it something like openclaw-prod. Copy the key (starts with sk-ant-).
4. In the Managed OpenClaw panel, navigate to Settings → LLM Provider, pick Anthropic, paste the key, and save.
5. Pick Claude Sonnet 4.6 as the default model. Switch to Opus 4.6 if you need stronger reasoning on harder tasks — but Opus is 5× the price of Sonnet, so save it for tasks where capability genuinely matters.

OpenAI GPT

1. Sign up at platform.openai.com
2. Go to Settings → Billing and add a payment method.
3. Go to API Keys → Create new secret key. Copy the key.
4. In the OpenClaw panel, pick OpenAI as the provider and paste the key.
5. Pick GPT-5.4 Mini for cheap reliable work, or GPT-5.4 (the full model) for harder tasks. Mini is around 10× cheaper and covers 80% of agent workloads adequately.

Google Gemini

1. Go to ai.google.dev and sign in with your Google account.
2. Click Get API key. Gemini has a generous free tier you can use to test before adding billing.
3. Copy the key and add it to OpenClaw with provider set to Google.
4. Pick Gemini 3 Flash — it's the cheapest reliable LLM in 2026 at $0.07/$0.30 per million tokens. For long-context work (document Q&A across hundreds of pages), use Gemini 3.1 Pro.

DeepSeek V3.2

1. Sign up at platform.deepseek.com
2. Add billing — DeepSeek requires a small minimum top-up (around $2).
3. Create an API key and copy it.
4. In OpenClaw, pick DeepSeek as the provider and paste the key.
5. Pick DeepSeek V3.2 — very strong on code tasks at $0.28/$1.10 per million. Good budget pick if your agent does a lot of code generation or technical reasoning.

Connecting OpenClaw to Telegram

Telegram is the most popular OpenClaw chat surface as of 2026 — partly because it's free and free of phone-number friction, partly because the OpenClaw Telegram skill is the most mature integration. Setup takes ~5 minutes.

Step 1: Create a Telegram bot via BotFather

1. Open Telegram and search for @BotFather (verified blue checkmark)
2. Send /newbot
3. BotFather asks for a display name (e.g., "My OpenClaw Agent") and a username (must end in bot, e.g., my_openclaw_bot)
4. BotFather replies with your bot token — a long string like 123456789:ABC...XYZ. Treat it like a password. Anyone with this token can impersonate your bot.

Step 2: Add the token to the Managed OpenClaw panel

In the panel, navigate to Integrations → Telegram → Add token and paste the bot token. The panel validates the token by calling Telegram's API — you'll know within a few seconds if it's valid.

Step 3: Get your Telegram user ID

OpenClaw needs to know which Telegram user is authorised to talk to it — otherwise anyone who finds your bot could send it commands. Get your numeric user ID:

1. Search for @userinfobot on Telegram
2. Send /start
3. It replies with your user ID (numeric, like 123456789)

Step 4: Authorise yourself

In the Managed OpenClaw panel, go to Integrations → Telegram → Authorised users and add your Telegram user ID. If you want multiple people to use the bot (e.g., your partner or teammates), add their user IDs to the same list.

Step 5: Say hi to your bot

In Telegram, search for the username you gave your bot, open the chat, and send:

/start

OpenClaw should respond within a few seconds. Try:

What can you do?

The agent will list its installed skills. Congratulations — your agent is now reachable from your phone, your laptop, anywhere Telegram works.

Connecting OpenClaw to Discord

Discord follows almost exactly the same pattern as Telegram — create a bot at Discord's developer portal, get a token, paste it into the panel, authorise yourself. Best if you live in Discord servers and prefer your agent there.

Step 1: Create a Discord application

1. Go to discord.com/developers/applications and click New Application
2. Name it whatever you want ("OpenClaw Agent" is fine)
3. In the sidebar, click Bot. Click Add Bot → Yes, do it!
4. Under Token, click Reset Token and copy the token that appears. This is the only time you'll see it — if you lose it, you have to reset and update the token in the OpenClaw panel.

Step 2: Configure bot permissions

Under Bot → Privileged Gateway Intents, enable Message Content Intent. Without this, your bot can't read messages and won't respond to anything.

Step 3: Add the token to the Managed OpenClaw panel

In the panel, navigate to Integrations → Discord → Add token and paste. The panel validates against Discord's API.

Step 4: Invite the bot to a server

Back in the Discord developer portal, go to OAuth2 → URL Generator. Check bot under Scopes, then in the permissions box check at minimum Send Messages and Read Message History. Copy the generated URL, paste it in your browser, and pick a Discord server you own to invite the bot to.

Step 5: Authorise yourself

You need your Discord user ID (a numeric snowflake). Enable Developer Mode in Discord settings (Advanced → Developer Mode), then right-click your username anywhere and pick Copy User ID. Add this ID to the panel's Authorised users list under Discord integration.

Step 6: Test it

In your Discord server, mention the bot or DM it: @OpenClaw Agent what can you do?. You should see a reply within a few seconds. If not, see the troubleshooting section.

Slack, Signal, iMessage, and WhatsApp follow the same general pattern — find each platform's developer portal, create a bot, get a token, paste it in the OpenClaw panel. Each has its own integration page in the panel.

Security on Managed OpenClaw

Most of OpenClaw's bad press in early 2026 came from researchers running it on their personal computer. The Cisco, Wired, and Immersive Labs writeups all describe what an agent could do on the machine it runs on — read local files, intercept browser sessions, hit your password manager, exfiltrate documents. On your laptop, the machine OpenClaw lives on is also the machine that holds your entire digital life. That's the threat surface.

On Hostinger Managed OpenClaw, the machine OpenClaw lives on is a container with nothing in it except OpenClaw and whatever you intentionally connect to it. Your laptop, your password manager, your tax returns, your browser sessions — none of those are reachable. The blast radius of a compromised agent shrinks to "whatever I put inside the container," which for most users is just an LLM API key and a Telegram bot token.

What Hostinger handles automatically on Managed OpenClaw:

• Container isolation — your agent can't see other customers' data or the host system
• OS patches — security updates land within hours of release
• Firewall — only the OpenClaw web UI is exposed, behind HTTPS
• SSL certificate — Let's Encrypt cert managed and renewed automatically
• DDoS mitigation — Hostinger's CDN/edge handles volumetric attacks
• Daily backups — automatic snapshots of your config and agent state

What you still need to handle yourself:

• Admin password hygiene — use a password manager, don't reuse passwords across services
• 2FA on your Hostinger account — enable it under your Hostinger dashboard settings
• Treat API keys as secrets — your LLM API key, Telegram bot token, Discord bot token all belong in the panel's encrypted secrets store, never in a plain text file or shared in chat
• Authorised-users lists — always set them on Telegram, Discord, etc., so random strangers can't message your bot
• Skill review — third-party OpenClaw skills can request broad permissions (filesystem access, shell execution, outbound network). Review what each skill needs before installing, and disable any skill you didn't intentionally install

Troubleshooting common errors

"Invalid API key" when adding LLM provider

Most often: your provider account doesn't have billing enabled yet, or the credit balance is zero. Anthropic, OpenAI, and DeepSeek all require a non-zero balance before keys work. Add $5 of credit and the key starts working immediately. If you're sure billing is set up, regenerate the key — keys created during account setup sometimes don't activate until billing is confirmed.

Panel says "Agent unreachable" or stuck on "Provisioning"

Wait a few more minutes — Hostinger's first-time provisioning can occasionally take up to 5 minutes on busy days. If it's been more than 10 minutes, contact Hostinger support directly from the panel (chat icon, bottom right). They can manually re-provision the container.

OpenClaw responds but very slowly

Two causes: (1) you picked a slow model — Claude Opus 4.6 is noticeably slower than Sonnet 4.6, which is slower than the smaller models. Switch to Sonnet 4.6 or GPT-5.4 Mini in the panel; the change applies immediately. (2) Your LLM provider's API is under load — check the provider's status page. Anthropic and OpenAI both publish public status pages.

Telegram bot doesn't respond after /start

Almost always one of these three: (a) bot token wasn't saved properly — re-enter it in the panel, (b) your user ID isn't in the Authorised users list — double-check it's the right numeric ID, (c) you accidentally created two bots and gave OpenClaw the wrong token. Check the panel's Telegram logs (visible in the Integrations → Telegram view) for the actual error message.

Discord bot doesn't respond to mentions

Most common cause: Message Content Intent wasn't enabled in the Discord developer portal. Without it, Discord doesn't send your bot the text of messages — it only sees that something was posted. Go back to the developer portal, enable the intent under Bot → Privileged Gateway Intents, and restart your agent from the panel.

API costs are way higher than the calculator predicted

Two common causes: (1) you're using a skill that builds long conversation context (e.g., "summarise this 30-page PDF") — these burn input tokens fast and a $1-$5 task is normal; (2) your agent has many skills enabled, each adding to the system prompt on every request — disable unused skills in the panel to slim the system prompt. Check actual usage in your LLM provider's dashboard for the exact breakdown.

"Persona changed but my agent acts the same"

The persona switch needs an agent restart to fully apply. In the panel, click Restart agent after switching personas. Existing conversation history is preserved.

Frequently asked questions

// Related Tools & Guides