A ready-to-customize starter kit for running your own AI assistant with OpenClaw.
- A personal AI assistant accessible via Discord, Telegram, or web
- Long-term memory across sessions
- Customizable personality and skills
- Free-tier model support (Groq, Cerebras, OpenRouter)
npm install -g openclawopenclaw onboardThis walks you through:
- Choosing your AI provider (Anthropic, OpenAI, Groq, etc.)
- Connecting Discord/Telegram (optional)
- Setting up your workspace
Edit the files in your workspace (created during onboard):
~/.openclaw/workspaces/<your-workspace>/
├── AGENTS.md ← How your agent behaves (read this first)
├── SOUL.md ← Personality, name, tone, directives
├── USER.md ← Info about you (name, timezone, preferences)
├── TOOLS.md ← Local notes (SSH hosts, device names, etc.)
├── HEARTBEAT.md ← Periodic tasks checklist
├── MEMORY.md ← Long-term memory (auto-maintained)
└── memory/ ← Daily session logs
openclaw gateway start# My Assistant
**Name:** Atlas
**Role:** Personal AI assistant
**Personality:** Friendly, concise, proactive
## What I Do
- Answer questions and research topics
- Manage tasks and reminders
- Monitor emails and calendar
- Help with coding and writing
## What I Don't Do
- Share private data with anyone
- Run destructive commands without asking
- Send messages on your behalf without permission# About Me
- **Name:** Your Name
- **Timezone:** Your/Timezone
- **Notes:** Whatever helps your assistant understand you# Heartbeat Tasks
- [ ] Check for unread emails
- [ ] Review calendar for upcoming events
- [ ] Any pending reminders?- Go to Discord Developer Portal
- Create a New Application → Bot
- Copy the bot token
- Run:
openclaw wizard discord - Invite bot to your server using the OAuth2 URL
- Message @BotFather on Telegram
- Create a new bot, get the token
- Run:
openclaw wizard telegram
| Provider | Model | Cost | Notes |
|---|---|---|---|
| Groq | Llama 3.3 70B | Free | Fast, good quality |
| Cerebras | GPT-OSS 120B | Free | Large context |
| OpenRouter | Various free models | Free | Many options |
| Ollama | Local models | Free | Runs on your machine |
Set your model in the config:
openclaw config set defaultModel groq/llama-3.3-70b-versatileopenclaw status # Check everything is running
openclaw gateway start # Start the gateway daemon
openclaw gateway stop # Stop it
openclaw gateway restart # Restart after config changes
openclaw doctor # Diagnose issues
openclaw memory search "query" # Search your memory- Config changes need a gateway restart to take effect
- Memory files are your continuity — the agent reads them every session
- HEARTBEAT.md runs every 30 min — use it for recurring checks
- Skills extend your agent — browse ClawHub for community skills
Built with ❤️ by MedGate Holdings