Getting Started on nacre.sh: Your First OpenClaw Agent

nacre.sh getting started guides tend to assume you know what OpenClaw is and why you're here. This one starts from scratch. If you've heard about OpenClaw, decided you want a personal AI agent, and landed on nacre.sh as the way to run it — this guide gets you from zero to a working agent without assuming any prior technical knowledge.

What You're About to Build

By the end of this guide, you'll have a personal AI agent running 24/7 in the cloud, accessible from Telegram (or Discord, Slack, or WhatsApp — your choice). Your agent will be able to answer questions, execute tasks using skills from ClawHub, and remember previous conversations.

What Is OpenClaw?

OpenClaw is an open-source AI agent framework — the software that runs underneath your AI assistant. It's the engine that connects your LLM (like Claude or GPT-4o) to the outside world: your calendar, your email, the web, and your messaging apps. OpenClaw has grown to over 361,000 GitHub stars as of mid-2026, making it the most popular self-hosted AI agent framework available.

nacre.sh runs OpenClaw for you in the cloud, so you get all the capability without any of the server management.

Step 1: Get an LLM API Key

Your agent needs a brain. nacre.sh supports 290+ LLM providers, but for beginners, we recommend starting with one of:

• Anthropic Claude: Best overall reasoning, $0.003–0.015 per 1K tokens depending on model
• OpenAI GPT-4o: Great for general tasks, widely supported by skills
• DeepSeek: Excellent cost-to-performance ratio if you're watching your budget

Go to your chosen provider's website, create an account, and generate an API key. Keep it somewhere safe — you'll paste it into nacre.sh in a moment.

Step 2: Create Your nacre.sh Account

Visit nacre.sh/signup. Create an account, verify your email, and add a payment method. The $12/month plan covers everything you need to get started.

Step 3: Deploy Your First Instance

Click New Instance in the dashboard. Choose a region (pick the one closest to your location), give your instance a name, and paste your LLM API key when prompted. Click Deploy. In about 90 seconds, your OpenClaw agent is running in the cloud.

Step 4: Connect Telegram

In the dashboard, go to Settings → Channels → Telegram. You'll need a Telegram bot token:

1. Open Telegram, search for @BotFather
2. Send /newbot and follow the prompts (give your bot a name and username)
3. BotFather will give you a token like 1234567890:AAF...
4. Paste this token into the nacre.sh Telegram setup field
5. Click Connect

Now open Telegram, find your new bot, and send it a message. Your OpenClaw agent will respond.

Step 5: Add Your First Skill

Skills give your agent capabilities beyond basic conversation. From the dashboard, go to Skills → Browse ClawHub and install the Web Search skill. This allows your agent to search the internet for current information. Install it with one click.

Test it by sending your agent: "Search for the latest OpenClaw news."

What to Explore Next

Once you're comfortable with basic interactions, explore:

• Calendar skill: Connect Google Calendar for scheduling assistance
• Memory skill: Give your agent persistent long-term memory
• Email skill: Have your agent manage your inbox

Frequently Asked Questions

Does my agent learn and remember things?

Yes. OpenClaw includes a memory system that persists between conversations. You can also explicitly tell your agent to remember specific facts.

How much will this cost in LLM API fees?

Light usage (20–30 messages per day) typically costs $3–8/month in API fees with Claude or GPT-4o. DeepSeek users spend roughly 80% less.

Can I change LLM providers after setup?

Yes. Go to Settings → LLM Providers and update your API key at any time. The change takes effect immediately.