OpenClaw
OpenClaw is an open-source, self-hosted AI agent that runs inside messaging apps like WhatsApp, Telegram, and Slack, executing real tasks such as managing files, sending emails, and automating workflows with your own model API keys.
What is OpenClaw?
OpenClaw is an open-source autonomous AI agent stewarded by the OpenClaw Foundation, an independent 501(c)(3)[reference:7]. It runs on your own Mac, Windows, or Linux machine and connects to messaging platforms—WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Microsoft Teams, Matrix, and others—so you interact with it through chat apps instead of a separate interface[reference:8].
The distinction that matters most is that OpenClaw is an agent, not a chatbot. A chatbot responds with text. OpenClaw acts: it can manage files, control a browser, send emails, schedule tasks, and chain multi-step workflows without you specifying every step[reference:9]. The software is free and MIT-licensed; your costs come from the model providers you connect (OpenAI, Anthropic, or local models via Ollama) and the infrastructure you run it on[reference:10].
How does OpenClaw work?
Users install OpenClaw via npm or pnpm (Node.js 22+), complete an onboarding wizard to connect model API keys and messaging platforms, and optionally install skills that extend the agent's capabilities[reference:11]. Once running, you send instructions in a supported chat app. OpenClaw interprets intent using the configured language model, selects the appropriate tool or skill, executes the action on the host machine or connected services, and returns the result in the same chat thread. For multi-step requests—such as researching competitors and emailing a summary—it chains actions automatically[reference:12].
Because OpenClaw is self-hosted, the agent remains active even when your personal device is off, provided the host machine stays online. Tool access can be restricted via profiles (minimal, coding, messaging, full), allow/deny lists, and per-provider policies[reference:13]. Sandbox isolation and elevated-mode approvals are available but require explicit configuration.
Use Cases
- Delegating inbox triage, email drafting, and calendar management from a chat thread
- Automating web research and having the agent compile and send summaries
- Running shell commands, managing files, and applying code patches through conversational instructions
- Connecting to smart home devices, productivity tools, and 50+ third-party services via skills[reference:14]
- Setting up scheduled or proactive workflows that run without an active chat session
Pricing
Pricing model: Free
| Plan | Price | Billing | Included usage | Notes |
|---|---|---|---|---|
| Self-Hosted (Open Source) | Free | — | Full source code under MIT license. You bring your own model API keys (OpenAI, Anthropic, or local models via Ollama) and are responsible for hosting and maintenance. | — |
| Managed Hosting (Third-Party) | Varies | — | Third-party services such as OpenClawHQ start at $49/month with AI usage included, but these are not first-party offerings from the OpenClaw Foundation. | — |
Common Limitations
- Security responsibility · OpenClaw executes real actions with system-level access. Sandbox isolation is disabled by default and must be explicitly enabled[reference:1]. Users are responsible for reviewing skills, restricting tool policies, and isolating the runtime environment. Microsoft has stated that OpenClaw is 'not appropriate to run on a standard personal or enterprise workstation' without additional controls[reference:2].
- Technical setup · Requires Node.js 22 or higher, a machine that stays on for 24/7 availability, and manual configuration of model providers, chat platform bridges, and optional skills. Setup is performed via CLI or desktop apps; there is no hosted first-party service.[reference:3]
- API cost variability · Every reply and tool call consumes model API tokens. Costs depend on the chosen provider, model, and usage volume. Claude Pro/Max consumer subscriptions cannot be used; Anthropic's terms require pay-as-you-go API keys instead[reference:4].
- Skill supply chain · Community-contributed skills and plugins extend the agent but are not uniformly audited. The official FAQ advises reviewing skills before installation and keeping the runtime updated[reference:5].
- Claude subscription restriction · Using an Anthropic Claude Pro or Max subscription as the model backend violates Anthropic's Terms of Service. OpenClaw users must configure pay-as-you-go API keys for Claude models[reference:6].
Key Features
- Runs inside WhatsApp, Telegram, Discord, Slack, Signal, iMessage, and 20+ messaging platforms
- Executes real tasks on the host machine: file management, browser control, shell commands, email, and calendar operations
- 100+ preconfigured AgentSkills plus support for community plugins
- Model-agnostic: supports Anthropic Claude, OpenAI GPT models, and local models via Ollama
- Tool policy profiles (minimal, coding, messaging, full) with per-agent and per-provider allow/deny controls
OpenClaw vs. Claude and Gemini
This comparison focuses on documented product positioning, execution model, and pricing structure. It does not rank model quality or subjective output preference. OpenClaw is self-hosted software; Claude and Gemini are managed consumer services, so the comparison highlights structural differences rather than feature parity.
| Tool | Positioning | Key differences | Best fit | Pricing note |
|---|---|---|---|---|
| OpenClaw Official homepage Official FAQ Tool documentation | An open-source, self-hosted autonomous agent that executes tasks inside messaging apps using your own model API keys. | Runs on your hardware, acts rather than only responds, supports 20+ messaging platforms, and gives you full control over tool policies and model providers. Security hardening and API cost management are your responsibility. | Users who want a self-hosted, model-agnostic agent that executes real tasks inside existing chat apps, and who are prepared to manage security, uptime, and API costs themselves. | Software is free (MIT). Model API costs are separate and usage-dependent, typically $10–30/month for light use and $100–150+/month for heavy automation. No first-party hosted tier. |
| Claude Official pricing | Anthropic's managed AI assistant for work, chat, coding, and document creation. | Cloud-hosted with no self-hosting option. Strong document, coding, and Artifacts capabilities, but action execution is limited to connected tools and does not include system-level access to your local machine by default. | Users who want a polished, managed assistant for writing, analysis, and coding, and prefer a subscription model over self-hosting and API cost management. | Free plan available. Pro is $20/month (or $200/year when billed annually). Paid plans include usage limits that vary by model and feature[reference:22]. |
| Gemini Official overview | Google's managed multimodal AI assistant integrated with Google's ecosystem. | Cloud-hosted with deep integration into Google products. Supports multimodal input and research features, but does not provide self-hosted system-level task execution on your own hardware. | Users who want a managed multimodal assistant closely connected to Google services and prefer a consumer subscription or free tier over self-hosting. | Free access is available, with paid Google AI plans offering higher limits and features. Regional pricing and availability vary. |
Comparison checked against official product information on 2026-11-08. Features and pricing may change.
Frequently Asked Questions
Is OpenClaw free?
The OpenClaw software is free and open-source under the MIT license. There is no first-party subscription or hosted tier. However, you pay for the AI model APIs you connect. The official FAQ estimates light use at $10–30/month, typical use at $30–70/month, and heavy automation at $100–150+/month[reference:15].
Which messaging apps does OpenClaw support?
OpenClaw bridges WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Microsoft Teams, Matrix, and more. It supports direct messages and group conversations with mention-based activation[reference:16].
Can I use OpenClaw with local models?
Yes. OpenClaw is model-agnostic and supports local models via Ollama. For maximum privacy, the documentation describes configuring memory search to use a local provider for zero cloud dependencies[reference:17][reference:18].
Can I use my Claude Pro or Max subscription with OpenClaw?
No. Anthropic's Terms of Service prohibit using consumer Claude Pro/Max subscriptions through third-party tools in this way. OpenClaw users must configure pay-as-you-go API keys for Claude models[reference:19].
What are the security considerations of self-hosting OpenClaw?
OpenClaw has system-level access and executes real actions. Sandbox isolation is disabled by default and must be explicitly enabled. The official documentation recommends reviewing skills before installation, using tool allow/deny policies, and following the security hardening guide. Microsoft has cautioned that OpenClaw should not be run on a standard workstation without additional isolation[reference:20][reference:21].
Alternative Tools
-
Manus Manus is an autonomous general-purpose AI agent that plans and executes multi-step tasks in a cloud sandbox, producing research reports, websites, slides, data analyses, and working prototypes rather than conversational answers. -
Grok Bot Grok Bot is xAI's AI teammate that runs tasks asynchronously in a cloud computer with browser, file system, and terminal access, available exclusively through The X Subscription's Premium tier and above. -
Browser Use Browser Use is an MIT-licensed open-source AI browser agent (Python and TypeScript) that lets LLMs control a browser, paired with a usage-based cloud offering at $0.02 per browser-hour for managed stealth browsers, CAPTCHA solving, and a hosted agent API. -
Devin Devin is an autonomous AI software engineer that plans, writes, tests, and ships code in asynchronous cloud sessions. It suits scoped backlog tasks and feature implementation on bounded codebases, but requires human supervision, struggles on large repositories, and carries a cost profile that is difficult to predict beyond included quotas. -
Genspark Genspark is an all-in-one AI workspace that combines autonomous agents, deep research, AI Slides, Sheets, and Docs, and 70+ models to turn prompts into finished work products—not just chat replies. -
Replit Replit is a browser-based IDE with an autonomous Agent 3 that builds, tests, and deploys apps from natural language, running up to 200 minutes without supervision. Core at $20/month is the real entry point; Starter is a demo tier. Effort-based billing, documented agent reliability issues, and a pattern of billing and support complaints are material risks for production use.