OpenClaw as a Personal AI Assistant: Full Setup

Building OpenClaw as a personal AI assistant is the goal that brings most people to the platform. A well-configured personal assistant knows your schedule, manages your communications, researches topics for you, and operates continuously even when you're not at your desk. Here's how to configure one that actually works in daily life.

Defining What "Personal Assistant" Means

Before starting, define what your assistant needs to do. Common functions:

• Memory: Remembers your preferences, important dates, ongoing projects
• Calendar management: Reads your schedule, can book meetings
• Email: Summarizes inbox, drafts replies, flags urgent messages
• Research: Answers questions using web search
• Daily briefing: Morning summary of calendar, weather, news, flagged items
• Task management: Tracks to-dos, reminds you of deadlines

Don't try to add everything at once. Start with memory + calendar + web search, and add capabilities after each proves reliable.

Step 1: Deploy Your Instance

Use nacre.sh for the fastest path. Alternatively, set up a self-hosted instance on a VPS or Mac Mini. Your assistant needs to be always-on to be genuinely useful.

Step 2: Craft Your System Prompt

The system prompt defines your assistant's personality and operating instructions. Keep it concise but specific:

You are a personal AI assistant. Your name is [Name]. 
Your user is [Your Name], a [your role].
Preferred communication: concise, direct, no filler.
When uncertain, ask one clarifying question rather than guessing.
Always check the calendar before scheduling anything.
Flag emails from [important domains] as high priority.

Step 3: Install Core Skills

python -m openclaw skill install memory        # Persistent memory
python -m openclaw skill install google-calendar  # Calendar integration
python -m openclaw skill install gmail-manager    # Email management
python -m openclaw skill install web-search      # Research capability
python -m openclaw skill install daily-briefing  # Morning summaries

Step 4: Connect Your Primary Channel

For a personal assistant, Telegram is ideal — it's on your phone, supports rich formatting, and is reliable. Connect it via @BotFather as described in the Telegram setup guide.

Step 5: Configure Memory

Explicitly instruct your assistant in early conversations about important context:

• "My name is Sarah. I'm a product manager at TechCorp. I prefer bullet points in summaries."
• "Remember that I have a standing meeting every Tuesday at 2 PM."
• "Always check weather for San Francisco when I ask about weather."

Your assistant stores these in its memory system and applies them to future conversations.

Step 6: Set Up the Daily Briefing

In openclaw.json, configure the scheduler:

{
  "scheduler": {
    "tasks": [{
      "name": "morning_briefing",
      "cron": "0 7 * * 1-5",
      "action": "run_skill",
      "skill": "daily-briefing",
      "output_channel": "telegram"
    }]
  }
}

Making It Work Long-Term

A personal assistant improves with use as it accumulates memory. The first week requires the most active feedback — correct mistakes, add preferences, refine the system prompt. After 2–4 weeks of daily use, most people find their assistant has become genuinely indispensable.

Frequently Asked Questions

How do I prevent my assistant from being too chatty?

Include "Be concise. Limit responses to 3–5 sentences unless detail is explicitly requested" in your system prompt. LLMs respond well to explicit brevity instructions.

Can my assistant remember information between different devices?

Yes. Your assistant's memory is stored server-side (in ~/.openclaw/memory.db or in nacre.sh's managed storage), so it's accessible from any channel on any device.

What's the biggest mistake people make setting up a personal assistant?

Adding too many skills too fast. Each skill adds complexity and potential for the LLM to choose the wrong tool. Add skills one at a time, verify each works, then add the next.