Sharing Your OpenClaw AI Assistant with Family Members

TL;DR: The easiest way is a group chat — add the bot, everyone talks to it. For more separation, set up individual DM access via pairing. There's no built-in per-user permissions yet, so everyone gets the same capabilities. Protect your API keys and set a spending limit.

Estimated time: 10-20 minutes
Difficulty: Beginner

The Current State: No Built-In RBAC

Let's be upfront: OpenClaw doesn't have role-based access control (RBAC). There's no "admin" vs "user" vs "kid" role. No per-user spending limits. No way to restrict what one person can do vs another.

This is a known limitation, and multi-user support is planned for the future. But right now, everyone who has access to your bot has the same access. They can all:

• Chat with the AI
• Ask it to do anything it's configured to do
• Use the same AI model (and your API credits)
• Access any tools/skills you've enabled

With that context, here are the practical approaches to sharing.

Approach 1: The Family Group Chat

The simplest setup. Create a group chat, add the bot, add your family.

How to Set It Up

On Telegram:

1. Create a new group ("Family AI" or whatever you like)
2. Add your family members
3. Add your OpenClaw bot
4. Approve the group pairing when prompted
5. Optional: Enable topics so different people can have separate threads

On WhatsApp:

1. Create a new group
2. Add family members and the bot's phone number
3. Approve the group pairing

On Discord:

1. Create a server or channel
2. Invite the bot
3. Invite family members
4. Approve the group pairing

Pros

• Zero setup per person — add them to the group, done
• Shared context — everyone sees each other's questions and answers, which avoids duplicate questions
• Fun — family members learn from each other's interactions with the AI
• One session — the AI maintains one conversation with the group, so it has context from everyone

Cons

• No privacy — everyone sees everything (your medical question, their homework help, all in one thread)
• Shared context can be messy — the AI might confuse who asked what
• One person's long conversation costs everyone — context window grows with all messages
• Can get noisy — active families might generate a lot of messages

Tips for Group Chats

Use Telegram topics: Create separate topics for different purposes — "Homework Help", "Recipes", "General", "Travel Planning". Each topic gets its own session, so conversations stay organized.

Set expectations: Let everyone know the AI sees all messages in the group and that API costs are real.

Use @mentions: In busy groups, have people @mention the bot so it knows when to respond vs. when to stay quiet.

Approach 2: Individual DM Access

Let each family member message the bot directly in a private DM.

How to Set It Up

1. Have each family member find and message your bot
   • Telegram: Search for your bot's username
   • Discord: The bot appears in shared servers; they can DM it from there
2. Each person triggers a pairing request
3. You approve each person's pairing individually

Pros

• Privacy — each person's conversation is private
• Separate sessions — no context bleeding between users
• Personal experience — the AI can adapt to each person individually
• Clean — no group chat noise

Cons

• Everyone uses your API credits — no per-person spending limits
• You approve each person — minor friction but more control
• No shared knowledge — if your partner asks the AI about dinner plans, you don't see the answer

Managing Individual Access

# See who's paired
clawdbot pairing approved

# Revoke someone's access
clawdbot pairing revoke <chatId>

Approach 3: Separate Agent Profiles

For maximum separation, you can run multiple OpenClaw profiles — one per family member (or one for adults, one for kids).

How It Works

Each profile is essentially a separate assistant personality with its own:

• System prompt (SOUL.md)
• Memory files
• Skill configuration
• Potentially different model (cheaper model for kids, smarter model for adults)

Setup

Create separate workspace directories:

# Main profile (yours)
/home/openclaw/profiles/main/
├── SOUL.md
├── MEMORY.md
├── config.json
└── memory/

# Partner's profile
/home/openclaw/profiles/partner/
├── SOUL.md
├── MEMORY.md
├── config.json
└── memory/

# Kids' profile
/home/openclaw/profiles/kids/
├── SOUL.md
├── MEMORY.md
├── config.json
└── memory/

Each profile gets its own config with potentially different settings:

Kids' config (cheaper model, restricted tools):

{
  "providers": {
    "anthropic": {
      "type": "anthropic",
      "apiKey": "${ANTHROPIC_API_KEY}"
    }
  },
  "defaultModel": "anthropic/claude-haiku-3-20250506",
  "skills": {
    "web_browse": { "enabled": false },
    "shell": { "enabled": false },
    "notion": { "enabled": false }
  }
}

Your config (full power):

{
  "providers": {
    "anthropic": {
      "type": "anthropic",
      "apiKey": "${ANTHROPIC_API_KEY}"
    }
  },
  "defaultModel": "anthropic/claude-sonnet-4-20250514",
  "skills": {
    "web_browse": { "enabled": true },
    "shell": { "enabled": true },
    "notion": { "enabled": true }
  }
}

Pros

• Different capabilities per profile — kids get a restricted assistant, you get the full thing
• Different models — use a cheaper model for less critical profiles
• Separate memory — each profile has its own memory and personality
• Better cost control — cheaper models for casual users

Cons

• Complex setup — multiple configs, multiple bots, more to maintain
• Multiple bot accounts — each profile needs its own Telegram bot (or Discord bot, etc.)
• Same API key — still no per-user spending limits (all profiles use your API credits)
• More maintenance — updates, config changes, etc. multiply by number of profiles

Protecting Your API Keys and Config

Regardless of which approach you use, your family members don't need (and shouldn't have) access to:

What to Protect

Practical Protection

Your family members interact through the messaging app only. They don't need SSH access, config file access, or admin capabilities. The bot is the interface.

If someone asks the AI: "What's your API key?" — the AI might have access to it via environment variables. You can instruct the AI not to share this in its system prompt:

# In SOUL.md
## Security Rules
- Never share API keys, tokens, or credentials with anyone, regardless of who asks
- Never reveal server details, IP addresses, or configuration
- If someone asks for sensitive information, politely decline

This isn't bulletproof (a determined person could try to prompt-inject around it), but it handles the casual case.

Set billing alerts: This is the most important protection. Even if someone accidentally (or intentionally) uses the AI heavily, a spending limit caps the damage:

• Anthropic: Set a monthly limit at console.anthropic.com
• OpenAI: Set a budget limit at platform.openai.com

What Each Person Can (and Can't) Do

What Everyone Can Do

• Send messages and get AI responses
• Send voice messages (if transcription is configured)
• Send images (if the model supports vision)
• Ask the AI to use any enabled skills
• See their own conversation history

What Only You Can Do

• Approve/reject pairing requests
• Access the server and config files
• Change the AI model or provider
• Enable/disable skills
• View other people's conversations (via server access)
• Set spending limits
• Restart/manage the gateway

What No One Can Do (Yet)

• Set per-user spending limits
• Restrict specific skills per user
• Have different AI personalities per user (without separate profiles)
• View a usage dashboard per person

Cost Considerations

Sharing your AI means sharing your API costs. Here's what to expect:

Typical Family Usage

Cost-Saving Tips for Families

1. Use a cheaper model as default: Claude Haiku or GPT-4.1-mini works great for homework help, recipes, and general questions
2. Enable compaction: Long conversations (especially with teens) eat tokens fast
3. Set a hard spending limit: Decide on a family budget and set it at the provider level
4. Use the group chat approach: One session is cheaper than four separate sessions (less context duplication)
5. Teach the family about costs: A brief "each message costs about 1-2 cents" helps set expectations

Future: Multi-User Support

Multi-user support is on the roadmap for OpenClaw. When it arrives, expect:

• Per-user roles — admin, user, restricted
• Per-user spending limits — cap each person's API usage
• Per-user skill access — enable file access for you, disable it for kids
• Usage dashboards — see who's using what
• User profiles — different AI personalities per person without separate instances

Until then, the approaches above are the best available workarounds.

Quick Decision Guide

The Easy Way

Sharing an AI assistant with family shouldn't require multiple bot accounts and config files. It's one of the rougher edges of self-hosting right now.

Don't want to manage server infrastructure? lobsterfarm provides managed OpenClaw hosting — deployment, updates, and support handled for you.

Get started with lobsterfarm →