OpenClaw: The Open-Source AI Agent That Actually Does Things
root
Mar 16 2026
"A smart model with eyes and hands at a desk with a keyboard and mouse. You message it like a coworker and it does everything a person could do."
What Is OpenClaw?
OpenClaw is a free, open-source, autonomous AI agent that runs locally on your own machine — Mac, Windows, or Linux. Rather than being another chatbot you visit in a browser tab, OpenClaw lives on your computer (or a server you control) and acts as a persistent personal assistant that can truly do things on your behalf.
It was created by Austrian developer Peter Steinberger (founder of PSPDFKit) and originally launched in November 2025 under the name Clawdbot. After some trademark complications with Anthropic, it was briefly renamed Moltbot, and finally settled on OpenClaw in late January 2026. The project exploded in popularity almost immediately, earning over 60,000 GitHub stars in under 72 hours — one of the fastest-growing open-source repositories in GitHub history.
The core idea is simple but powerful: instead of you going to the AI, the AI lives with you, remembers your context, integrates with the tools you already use, and acts proactively even when you're not asking.
How It Works
OpenClaw operates through a local Gateway — a daemon (background service) that runs on your machine and acts as the control plane. This Gateway:
- Connects to your preferred messaging apps (WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Teams, and 20+ others)
- Integrates with an AI model of your choice (Claude via API, OpenAI GPT models, or local models via Ollama)
- Stores memory and context locally as Markdown files on your disk
- Executes skills — modular capabilities that extend what the agent can do
You interact with OpenClaw by simply sending a message in your chat app of choice. It processes your request, takes action, and responds — just like texting a coworker.
The Three Pillars
| Pillar | Description |
|---|---|
| Computer Access | Can read/write files, run shell commands, control your browser, and execute scripts |
| Persistent Memory | Remembers your preferences, projects, and context across all sessions |
| Heartbeat | Wakes up proactively on a schedule to run tasks without being prompted |
Key Features
🔒 Local-First & Private
Your data never leaves your machine unless you explicitly send it somewhere. Memory, files, and configuration are stored locally. You can even use fully local AI models via Ollama for a completely offline setup.
🛠️ Skills System
OpenClaw's functionality is extended through skills — directories containing a SKILL.md file with instructions and tool definitions. Skills can be:
- Bundled with the software
- Installed from ClawHub (the community skill registry)
- Custom-built for your own workflows
- Auto-generated by the agent itself when it needs a new capability
🤖 Model-Agnostic
Works with any major AI provider:
- Anthropic Claude (via API key)
- OpenAI GPT models
- Local models via Ollama (Llama, Mistral, etc.)
⚠️ Note: Using Claude Pro/Max subscriptions violates Anthropic's ToS — you must use pay-as-you-go API keys.
⏰ Proactive via Heartbeat
Unlike passive chatbots, OpenClaw can be configured to wake up at regular intervals, assess your situation, and act — sending you a morning briefing, checking flight statuses, or summarizing overnight emails without you asking.
Installation (Quick Start)
Requirements: Node.js ≥ 22, npm/pnpm/bun
# Install globally
npm install -g openclaw@latest
# Run the onboarding wizard (sets up the Gateway daemon)
openclaw onboard --install-daemon
The wizard guides you through:
1. Pairing with a messaging app (Telegram is the easiest starting point)
2. Connecting an AI model (via API key)
3. Configuring basic preferences and persona
Common Use Cases
1. 📬 Email & Calendar Management
OpenClaw monitors your inbox and calendar, takes appropriate action when someone wants to reschedule a meeting, checks your availability, and drafts confirmations — all without your involvement. You can ask it things like:
- "What's on my calendar this week?"
- "Reply to John and tell him I'm available Thursday afternoon."
- "Summarize the last 10 emails from my team."
2. 🌅 Daily Briefings
Configure a morning heartbeat that delivers a summary directly to WhatsApp or Telegram every day:
- Your schedule for the day
- Key emails that need attention
- Weather and commute info
- Relevant news based on your interests
3. 💻 Developer Workflows
This is where OpenClaw truly shines for technical users:
- Manage Claude Code / Codex sessions remotely from your phone
- Autonomously run tests, capture errors from Sentry, and open GitHub Pull Requests with fixes
- Refactor codebases, add TypeScript types, write unit tests
- Analyze PostgreSQL query plans and add indexes
- Deploy to Kubernetes clusters via openclaw deploy
4. 🌐 Browser Automation
OpenClaw controls a dedicated Chromium instance to:
- Fill out forms on your behalf (reimbursements, bookings, applications)
- Scrape and extract data from websites
- Research and compile information from multiple sources
- Make restaurant reservations or purchases
Real example: Developer AJ Stuyvenberg had his OpenClaw agent negotiate $4,200 off a car purchase while he slept — it scraped dealer inventories, forwarded competing quotes, and got dealers to undercut each other.
5. 🤝 Team & Community Assistant
Organizations use OpenClaw as a shared assistant in Slack or Discord:
- Answer common questions using internal documentation
- Triage issues and route them to the right person
- Summarize long threads and meetings
- Post updates and reminders automatically
7. 📁 File & Knowledge Management
- Organize files based on content and naming conventions
- Sync notes from Obsidian with calendar and task context
- Build a "second brain" as you chat — information is extracted and stored automatically
- Generate summaries and reports from local documents
Security Considerations
OpenClaw's broad access to your system is its greatest strength — and its most significant risk. Be aware of the following before deploying it:
- Prompt Injection: Malicious instructions embedded in emails or web pages can be processed as legitimate commands. Use strong, frontier-grade models to reduce this risk.
- Skill Vetting: Community skills are not rigorously reviewed. Treat skill installation like installing new code in a privileged environment — verify provenance.
- Over-Privileged Access: Only grant the permissions OpenClaw actually needs. Avoid giving it production secrets or sensitive credentials unless absolutely necessary.
- Sandbox It: For production or sensitive use, run OpenClaw in a sandboxed VM or Docker container with restricted network access.
One maintainer's own warning: "If you can't understand how to run a command line, this is far too dangerous of a project for you to use safely."
Pricing
OpenClaw itself is completely free (MIT license). Your only costs are AI model API calls:
| Usage Level | Estimated Monthly Cost |
|---|---|
| Light use | $10 – $30 |
| Typical use | $30 – $70 |
| Heavy automation | $100 – $150+ |
You can reduce costs by using local models via Ollama or the Brave Search free tier (2,000 requests/month).
Resources
| Resource | Link |
|---|---|
| Official Website | openclaw.ai |
| GitHub Repository | github.com/openclaw/openclaw |
| Community Discord | Available via the official site |
| Skill Registry (ClawHub) | Built into the CLI |
| Documentation | openclaw.ai/docs |
Summary
OpenClaw represents a meaningful shift in how personal AI assistance works. Instead of a passive chatbot you have to visit, it's a proactive, local-first agent that lives in your existing messaging apps, remembers everything, and takes real action — autonomously, around the clock.
It's best suited for developers and technical power users who want maximum control over their data and workflows. For non-technical users, the setup is approachable, but the security model demands careful attention.
Whether you want to automate your inbox, manage a dev pipeline from your phone, or build a custom assistant for your team — OpenClaw is one of the most capable and flexible open-source tools available today.
"The future is already here — it's just running on Node.js and texting you on WhatsApp."
root
Just share your knowledge!