test-repo/openclaw-tips-tricks-guide.md
Matt Bruce 8f4f979c85 Add specialized agent profiles and documentation
New specialized agents:
- ios-dev: iOS/Swift/SwiftUI development
- web-dev: Next.js/React/web development
- research: Research and analysis tasks

Each agent has:
- Own agent.json config with specialized system prompt
- Auto-loaded relevant skills
- Domain-specific best practices and rules
- Proper model configuration

Also added:
- SPECIALIZED_AGENTS.md documentation
- CLI_README_TEMPLATE.md for reference
2026-02-21 19:31:38 -06:00

17 KiB

OpenClaw Tips, Tricks, and Usage Patterns

A comprehensive guide compiled from official documentation, community resources, and real-world examples.

Last Updated: February 21, 2026
Document Version: 1.0


Table of Contents

  1. Introduction
  2. Best Practices and Workflows
  3. Tips for Beginners
  4. Tips for Advanced Users
  5. Creative Use Cases
  6. Integration Examples
  7. Community Insights
  8. Troubleshooting Quick Reference
  9. Resources and Attribution

Introduction

OpenClaw is a self-hosted AI assistant gateway that connects your favorite chat apps—WhatsApp, Telegram, Discord, iMessage, Slack, and more—to AI coding agents. It runs on your own hardware, giving you complete control over your data while providing always-available AI assistance.

Key Differentiators

  • Self-hosted: Runs on your hardware, your rules
  • Multi-channel: One Gateway serves multiple messaging platforms simultaneously
  • Agent-native: Built for coding agents with tool use, sessions, memory, and multi-agent routing
  • Open source: MIT licensed, community-driven

Best Practices and Workflows

1. Heartbeat vs Cron: When to Use Each

Understanding when to use heartbeat checks versus cron jobs is crucial for efficient automation.

Use Case Recommended Why
Check inbox every 30 min Heartbeat Batches with other checks, context-aware
Send daily report at 9am sharp Cron (isolated) Exact timing needed
Monitor calendar for upcoming events Heartbeat Natural fit for periodic awareness
Run weekly deep analysis Cron (isolated) Standalone task, can use different model
Remind me in 20 minutes Cron (main, --at) One-shot with precise timing
Background project health check Heartbeat Piggybacks on existing cycle

Decision Flowchart:

Does the task need to run at an EXACT time?
  YES → Use cron
  NO  → Continue...

Does the task need isolation from main session?
  YES → Use cron (isolated)
  NO  → Continue...

Can this task be batched with other periodic checks?
  YES → Use heartbeat (add to HEARTBEAT.md)
  NO  → Use cron

Source: OpenClaw Documentation - Cron vs Heartbeat

2. Memory Management Best Practices

OpenClaw memory is plain Markdown in the agent workspace. Files are the source of truth.

Default Memory Structure:

  • memory/YYYY-MM-DD.md - Daily log (append-only), read today + yesterday at session start
  • MEMORY.md - Curated long-term memory (load only in main, private sessions)

When to Write Memory:

  • Decisions, preferences, and durable facts → MEMORY.md
  • Day-to-day notes and running context → memory/YYYY-MM-DD.md
  • When someone says "remember this" → Write it immediately (don't keep in RAM)
  • If you want something to stick, ask the bot to write it

Pro Tip: Memory is limited—if you want to remember something, WRITE IT TO A FILE. "Mental notes" don't survive session restarts. Files do.

Source: OpenClaw Documentation - Memory

3. Secure DM Mode for Multi-User Setups

Security Warning: If your agent can receive DMs from multiple people, enable secure DM mode to prevent context leakage between users.

The Problem:

  • Alice messages about a private topic (e.g., medical appointment)
  • Bob asks "What were we talking about?"
  • With default settings, both share the same session context

The Fix:

{
  session: {
    dmScope: "per-channel-peer",  // Isolate DM context per channel + sender
  },
}

When to Enable:

  • You have pairing approvals for more than one sender
  • You use a DM allowlist with multiple entries
  • You set dmPolicy: "open"
  • Multiple phone numbers or accounts can message your agent

Source: OpenClaw Documentation - Session Management

4. Configuration Organization

Minimal Config (Single User):

{
  agent: { workspace: "~/.openclaw/workspace" },
  channels: { whatsapp: { allowFrom: ["+15555550123"] } },
}

Recommended Starter:

{
  identity: {
    name: "Clawd",
    theme: "helpful assistant",
    emoji: "🦞",
  },
  agent: {
    workspace: "~/.openclaw/workspace",
    model: { primary: "anthropic/claude-sonnet-4-5" },
  },
  channels: {
    whatsapp: {
      allowFrom: ["+15555550123"],
      groups: { "*": { requireMention: true } },
    },
  },
}

Source: OpenClaw Documentation - Configuration Examples


Tips for Beginners

1. Getting Started Fast

  1. Install OpenClaw:

    npm install -g openclaw@latest
    
  2. Run the Onboarding Wizard:

    openclaw onboard --install-daemon
    
  3. Pair WhatsApp and Start Gateway:

    openclaw channels login
    openclaw gateway --port 18789
    

Time to Setup: Usually 5-15 minutes

Source: OpenClaw Documentation - Quick Start

2. Essential CLI Commands to Know

# Check status
openclaw status

# View and manage sessions
openclaw sessions --json
openclaw sessions --active 30  # Last 30 minutes

# Check gateway health
openclaw gateway status
openclaw doctor  # Diagnose issues

# View logs
openclaw logs --follow

# Send a test message
openclaw message send --to +1234567890 --message "Hello from OpenClaw"

# Talk to the assistant
openclaw agent --message "Ship checklist" --thinking high

3. First-Time Configuration Tips

  • Start simple: Use minimal config, then expand
  • Enable pairing: Unknown senders receive a pairing code before the bot processes their message
  • Use mention gating in groups: Require @openclaw mention to avoid spam
  • Set up daily memory files: Ask the agent to write important information to memory

4. Resetting Sessions

When conversations get off track:

  • Send /new or /reset to start a fresh session
  • Use /new <model> to switch models (e.g., /new opus)
  • Sessions auto-reset daily at 4:00 AM by default

Tips for Advanced Users

1. Multi-Agent Routing

Route inbound channels/accounts/peers to isolated agents with separate workspaces:

{
  agents: {
    defaults: { workspace: "~/.openclaw/workspace" },
    list: [
      { id: "coding", workspace: "~/.openclaw/coding-agent" },
      { id: "personal", workspace: "~/.openclaw/personal-agent" },
    ],
  },
  routing: {
    rules: [
      { channel: "telegram", account: "work", agent: "coding" },
      { channel: "whatsapp", agent: "personal" },
    ],
  },
}

Source: OpenClaw Documentation - Multi-Agent

2. Model Failover Configuration

Set up automatic fallbacks when primary models fail:

{
  agents: {
    defaults: {
      model: {
        primary: "anthropic/claude-sonnet-4-5",
        fallbacks: ["anthropic/claude-opus-4-6", "openai/gpt-5.2"],
      },
    },
  },
}

Source: OpenClaw Documentation - Model Failover

3. Advanced Cron Scheduling

Daily Morning Briefing:

openclaw cron add \
  --name "Morning briefing" \
  --cron "0 7 * * *" \
  --tz "America/New_York" \
  --session isolated \
  --message "Generate today's briefing: weather, calendar, top emails, news summary." \
  --model opus \
  --announce \
  --channel whatsapp \
  --to "+15551234567"

One-Shot Reminder:

openclaw cron add \
  --name "Meeting reminder" \
  --at "20m" \
  --session main \
  --system-event "Reminder: standup meeting starts in 10 minutes." \
  --wake now \
  --delete-after-run

Source: OpenClaw Documentation - Cron Jobs

4. Skills Development

Skills are AgentSkills-compatible folders with SKILL.md files. Load locations (in precedence order):

  1. <workspace>/skills (highest)
  2. ~/.openclaw/skills
  3. Bundled skills (lowest)

Install from ClawHub:

clawhub install <skill-slug>
clawhub update --all

Skill Format:

---
name: my-custom-skill
description: Does something useful
metadata:
  { "openclaw": { "requires": { "bins": ["node"], "env": ["API_KEY"] } } }
---

Instructions for the agent on how to use this skill...

Source: OpenClaw Documentation - Skills

5. Canvas and A2UI Integration

The Canvas panel provides a visual workspace for HTML/CSS/JS and A2UI:

# Present canvas
openclaw nodes canvas present --node <id>

# Navigate to content
openclaw nodes canvas navigate --node <id> --url "/"

# Evaluate JavaScript
openclaw nodes canvas eval --node <id> --js "document.title"

# Capture snapshot
openclaw nodes canvas snapshot --node <id>

# Push A2UI content
openclaw nodes canvas a2ui push --node <id> --text "Hello from A2UI"

Source: OpenClaw Documentation - Canvas

6. Voice Wake Configuration

Voice wake words are global (owned by the Gateway), stored at:

  • ~/.openclaw/settings/voicewake.json
{ "triggers": ["openclaw", "claude", "computer"], "updatedAtMs": 1730000000000 }

Each device keeps its own enabled/disabled toggle while sharing the global trigger list.

Source: OpenClaw Documentation - Voice Wake


Creative Use Cases

From the Community Showcase

1. PR Review → Telegram Feedback

By: @bangnokia
Flow: OpenCode finishes change → opens PR → OpenClaw reviews diff → replies in Telegram with suggestions and merge verdict

2. Wine Cellar Skill

By: @prades_maxime
Flow: Asked for local wine cellar skill → requests CSV export + storage location → builds/tests skill (962 bottles)

3. Tesco Shop Autopilot

By: @marchattonhere
Flow: Weekly meal plan → regulars → book delivery slot → confirm order (no APIs, just browser control)

4. SNAG Screenshot-to-Markdown

By: @am-will
Flow: Hotkey screen region → Gemini vision → instant Markdown in clipboard
GitHub: https://github.com/am-will/snag

5. Multi-Agent Content Factory

By: Community
Flow: Run multi-agent content pipeline in Discord — research, writing, and thumbnail agents in dedicated channels

Source: OpenClaw Showcase

From Awesome OpenClaw Use Cases

Social Media & Content

  • Daily Reddit Digest: Summarize curated subreddits based on preferences
  • Daily YouTube Digest: Get summaries of new videos from favorite channels
  • X Account Analysis: Qualitative analysis of your X/Twitter account
  • Multi-Source Tech News: Aggregate from 109+ sources (RSS, Twitter, GitHub, web search)

Productivity

  • Autonomous Project Management: Multi-agent projects using STATE.yaml pattern
  • Phone-Based Personal Assistant: Access AI via voice calls/SMS
  • Inbox De-clutter: Summarize newsletters and send digests
  • Personal CRM: Auto-discover and track contacts from email/calendar
  • Second Brain: Text anything to remember, search via Next.js dashboard

Infrastructure

  • Self-Healing Home Server: Always-on infrastructure agent with SSH access and automated cron jobs
  • n8n Workflow Orchestration: Delegate API calls to n8n workflows via webhooks

Source: Awesome OpenClaw Use Cases


Integration Examples

1. n8n Workflow Orchestration

Delegate API calls to n8n workflows—the agent never touches credentials:

# OpenClaw calls n8n webhook
openclaw agent --message "Trigger n8n workflow to add user to Mailchimp"

Benefits:

  • Visual workflow builder
  • Lockable integrations
  • No API keys in agent context

2. Browser Automation

# Check browser status
openclaw browser status

# Start browser
openclaw browser start --browser-profile openclaw

# Chrome extension relay (for taking over existing tabs)
openclaw browser profiles

3. Node Integration (iOS/Android)

Pair mobile nodes for camera, location, and screen capture:

# List paired nodes
openclaw nodes status

# Take camera snapshot
openclaw nodes camera_snap --node <id> --facing front

# Start screen recording
openclaw nodes screen_record --node <id> --duration 30s

# Get location
openclaw nodes location_get --node <id>

4. Webhook Integration

{
  hooks: {
    enabled: true,
    path: "/hooks",
    token: "shared-secret",
    presets: ["gmail"],
    mappings: [
      {
        id: "gmail-hook",
        match: { path: "gmail" },
        action: "agent",
        wakeMode: "now",
        name: "Gmail",
        deliver: true,
        channel: "last",
      },
    ],
  },
}

5. Custom Model Providers

{
  models: {
    providers: {
      "custom-proxy": {
        baseUrl: "http://localhost:4000/v1",
        apiKey: "LITELLM_KEY",
        api: "openai-responses",
        models: [
          {
            id: "llama-3.1-8b",
            name: "Llama 3.1 8B",
            contextWindow: 128000,
            maxTokens: 32000,
          },
        ],
      },
    },
  },
}

Community Insights

From the community and docs:

  • Recommended: Anthropic Pro/Max ($100/$200) + Opus 4.6 for long-context strength and prompt-injection resistance
  • Fallback chain: Claude Sonnet 4.5 → Claude Opus 4.6 → GPT-5.2

Security Best Practices

  1. Always use pairing for unknown senders (dmPolicy: "pairing")
  2. Approve with: openclaw pairing approve
  3. Audit regularly: openclaw security audit
  4. Review third-party skills before enabling (treat as untrusted code)
  5. Use sandboxed runs for untrusted inputs
  1. Telegram - Easy bot setup, great for beginners
  2. WhatsApp - Most personal, requires phone pairing
  3. Discord - Great for multi-user/group collaboration
  4. Slack - Enterprise/team workflows
  5. iMessage - Requires macOS (BlueBubbles recommended)
  6. Signal - Privacy-focused

Development Workflow Tips

  1. Use openclaw doctor to diagnose issues
  2. Check logs with openclaw logs --follow
  3. Test skills locally before deploying
  4. Use isolated sessions for heavy tasks
  5. Set up heartbeat for routine monitoring instead of multiple cron jobs

Troubleshooting Quick Reference

No Replies from Bot

openclaw status
openclaw channels status --probe
openclaw pairing list <channel>
openclaw logs --follow

Common causes:

  • Pairing pending for DM senders
  • Group mention gating (requireMention)
  • Channel/group allowlist mismatches

Gateway Won't Start

openclaw gateway status
openclaw doctor
openclaw logs --follow

Common causes:

  • gateway.mode not set to local
  • Port conflict (EADDRINUSE)
  • Non-loopback bind without auth

Cron/Heartbeat Not Running

openclaw cron status
openclaw cron list
openclaw system heartbeat last

Common causes:

  • Cron scheduler disabled
  • Outside active hours window
  • Invalid account ID for delivery target

Browser Tool Fails

openclaw browser status
openclaw browser start --browser-profile openclaw
openclaw doctor

Common causes:

  • Chrome not found at configured path
  • Extension relay not attached (for Chrome profile)

Source: OpenClaw Documentation - Troubleshooting


Resources and Attribution

Official Documentation

Community Resources

Key Contributors and Sources

  • OpenClaw Team: Core platform development and documentation
  • VoltAgent: Awesome OpenClaw Skills curation
  • @hesamsheikh: Awesome OpenClaw Use Cases compilation
  • Community Showcase Contributors: @bangnokia, @prades_maxime, @marchattonhere, @am-will, @kitze, and many others

Videos and Tutorials

  • Cherry Studio: AI productivity studio with OpenClaw integration
  • 1Panel: Linux server management with OpenClaw support
  • Umbrel: Home server OS with OpenClaw app
  • MimiClaw: Run OpenClaw on a $5 chip (no OS, no Node.js)

License and Usage

This document is a compilation of publicly available information from OpenClaw documentation and community resources. All source materials are attributed above.

OpenClaw itself is released under the MIT License.


Document compiled by OpenClaw Research Agent - February 2026

For the latest information, always refer to the official documentation at https://docs.openclaw.ai