The datasheet for every AI agent

Open source AI assistant that runs on your machine and manages email, calendar, and tasks via WhatsApp or Telegram

Visit OpenClaw Compare
From
$20 trial credit / monthly
Free trial
3 days
Pricing model
free
Free tier
Yes
Deployment
Self-host
Interface
CLI
Open source
Yes
Public API
Yes
Model / LLM
Model-agnostic — hosted Claude,…

Facts re-verified

Our verdict

OpenClaw is a genuinely open, self-hosted alternative to closed personal-assistant apps: MIT-licensed, installable in one command, and wired into 29 chat channels and 64 model providers so it can sit inside WhatsApp or Telegram rather than another silo.

The heartbeat scheduler and command-line agency make it feel closer to a real assistant than a chatbot. The catch is what happens after install. One reviewer spent roughly six hours getting it running on a VPS, watched it drop tasks with no dashboard to track them, and later saw the bill jump from about $20 to $500+ a month once Anthropic stopped covering third-party API usage. It rewards people willing to operate infrastructure, not people who want a finished product.

What is OpenClaw?

OpenClaw is an open source, MIT-licensed personal AI assistant that installs and runs on your own machine rather than a vendor's cloud. Built by Peter Steinberger and community contributors under the OpenClaw Foundation, it connects to chat apps you already use, including WhatsApp, Telegram, Discord, and Slack, so you talk to it the same way you'd message a person.

It's aimed at technical users comfortable with a command line who want an always-on assistant handling email, calendar, and routine admin without routing personal data through a third party's servers.

What does OpenClaw do?

OpenClaw installs with a single curl command or via npm, then runs as a persistent agent on macOS, Linux, or Windows. Once set up, it organizes your inbox, drafts and sends emails, manages your calendar, and can check you in for flights, all reachable from WhatsApp, Telegram, or any of the 29 supported chat channels rather than a dedicated app.

It stays proactive through a heartbeat configuration file, similar to a cron schedule, that determines how often it wakes to check its task queue and message you. It remembers context across conversations by writing notes to a markdown file, and it has its own command-line access, meaning it can write scripts, execute code, and install packages on its own to complete a task.

Under the hood it can route requests through 64 model and media providers, including Anthropic Claude, Google Gemini, OpenAI, and local runtimes like Ollama and LM Studio, plus inference gateways such as Amazon Bedrock and OpenRouter. Its 142 official plugins and the ClawHub public registry let users add or share reusable skills.

Beta companion apps for macOS and a Windows Hub extend it beyond the terminal, and it can also run on a VPS or cloud host such as Fly.io, Hetzner, or Azure for continuous uptime instead of depending on a laptop staying on.

How OpenClaw works

  1. Install with npm install -g openclaw@latest --allow-scripts=openclaw, then run openclaw onboard --install-daemon to set up the Gateway as a background service.
  2. Connect a chat app such as WhatsApp, Telegram, Discord, or Signal so you can message the assistant the way you'd message a person.
  3. Authenticate a model provider, for example openclaw models auth login --provider anthropic --method cli --set-default, to route requests through Claude, Gemini, GPT, xAI Grok, or a fully local model like Ollama.
  4. Set a heartbeat schedule so OpenClaw wakes periodically to check its task queue and proactively message you.
  5. Open the browser Control UI with openclaw dashboard (default http://127.0.0.1:18789/) for chat, config, and session history.

Key features

Chat-native interface
Works from WhatsApp, Telegram, Discord, Slack, Signal, iMessage, and 29 chat channels in total, so there's no separate app to open.
Always-on heartbeat
Uses a cron-like heartbeat config file to wake periodically, check its task queue, and message you proactively rather than only responding when asked.
Broad model routing
Connects to 64 model and media providers, including Anthropic Claude, Google Gemini, OpenAI, and local runtimes like Ollama, LM Studio, vLLM, and SGLang.
Command-line agency
Has its own command-line access to write scripts, execute code, and install packages on the fly rather than just returning text.
Plugin ecosystem via ClawHub
142 official plugins plus a public registry, ClawHub, for installing or publishing reusable skills.
Markdown-based memory
Recalls context about you across conversations by writing to a markdown file rather than a hosted vector database.

What are OpenClaw's use cases?

Inbox and calendar triage
A self-employed user connects OpenClaw to their email and calendar so it drafts replies, files messages, and surfaces what needs attention, all managed from a WhatsApp thread instead of a dashboard.
Automated flight check-in
A frequent traveler lets OpenClaw watch for upcoming flights and check them in automatically, since flight check-in is one of the tasks it's built to handle.
Self-hosted assistant on a home server or VPS
A developer runs OpenClaw on a spare machine or a VPS, such as Fly.io or Hetzner, so it stays always-on without depending on a laptop being open, using the heartbeat scheduler to check in periodically.
Extending capability through ClawHub plugins
A power user browses the ClawHub public registry to install or publish reusable skills, adding to the 142 official plugins already available instead of building integrations from scratch.

Who is OpenClaw for?

  • Technically comfortable individuals who want a self-hosted, always-on assistant for personal admin like email and calendar
  • Open source enthusiasts who prefer running AI tools on their own hardware instead of a vendor's cloud
  • Developers who want to script or extend an assistant via plugins and a public skill registry

Not forNon-technical users who want a polished, predictable-cost app: setup can take hours, there's no built-in task tracker, and monthly cost can swing from about $20 to $500+ depending on how model billing is routed.

What does OpenClaw integrate with?

  • WhatsApp
  • Telegram
  • Discord
  • Slack
  • Signal
  • iMessage
  • Matrix
  • IRC
  • Microsoft Teams
  • Google Chat
  • LINE
  • Mattermost
  • Feishu/Lark
  • Twitch
  • Zalo
  • Voice Call (Twilio/Telnyx/Plivo)
  • WebChat
  • Anthropic Claude API/CLI
  • Google Gemini
  • OpenAI/ChatGPT/Codex
  • xAI Grok
  • Mistral
  • DeepSeek
  • Moonshot/Kimi
  • Qwen
  • Cohere
  • Ollama
  • LM Studio
  • vLLM
  • SGLang
  • Amazon Bedrock
  • OpenRouter
  • LiteLLM
  • Groq
  • Cerebras

Why use OpenClaw?

  1. 142 official plugins and a public ClawHub registry for adding new skills
  2. 64 model and media provider integrations, including Claude, Gemini, GPT, and local Ollama models
  3. 29 supported chat channels so it fits inside WhatsApp, Telegram, Discord, or Slack instead of a new app
  4. Fully open source under MIT license, with source available on GitHub as a pnpm workspace

What are OpenClaw's pros and cons?

What's great

  • Fully open source (MIT license) with source on GitHub, so nothing is locked behind a proprietary vendor
  • Deep integration surface out of the box: 29 chat channels, 64 model providers, and 142 official plugins
  • Proactive by design via a heartbeat scheduler, rather than only responding when messaged

Watch-outs

  • One reviewer's setup took about 6 hours on a VPS, with no polished installer experience for non-technical users
  • No built-in task management view, and the reviewer reported it dropping some assigned tasks
  • Real-world cost can be unpredictable: one reviewer's bill went from about $20/month under a Claude subscription to $500+/month after Anthropic ended third-party tool coverage and billing moved to direct API rates
  • A commenter, not the reviewer, described OpenClaw's memory as breaking down with longer-term usage, based on secondhand reports rather than direct hands-on testing
  • Pricing is not centrally published by OpenClaw itself; the cost figures here come from a third-party blog post and a LinkedIn review rather than an official pricing page
  • There is no built-in task management system to see what has been assigned, per one reviewer's hands-on notes

OpenClaw pricing

from $20/mo (can rise to $500+/mo under direct API billing)

3-day free trial

  • Self-hosted (open source)

    Free (one-time)

    • MIT-licensed open source, free to run, fork, and modify forever
    • You pay only your own model-provider API costs and any hosting (home server or VPS)
  • Elestio hosted (managed)

    $20 trial credit / monthly

    • Free trial: $20 in credits valid for 3 days
    • Then hourly credit-based billing through Elestio on Netcup, Hetzner, AWS Lightsail, DigitalOcean, Vultr, Linode, Scaleway, or AWS
    • Tier 1 support included free; paid support plans available separately

See current pricing on openclaw.ai ↗Compare OpenClaw alternatives →

Frequently asked questions

Is OpenClaw free to use?
The software itself is open source and MIT licensed, so running it on your own machine costs nothing beyond your own hardware and any model API usage. Actual monthly cost depends on which model provider you connect and how billing is routed; one reviewer reported paying around $20/month under a Claude subscription, which rose to $500+/month after Anthropic ended third-party tool coverage and billing moved to direct API rates.
What platforms does OpenClaw run on?
It installs on macOS, Linux, and Windows via a single curl command or npm i -g openclaw, which also pulls in Node.js. Beta companion apps exist for macOS (macOS 15+) and a Windows Hub (Windows 10 20H2+ or Windows 11).
Can I try OpenClaw without installing it myself?
Elestio offers a hosted option with a free trial giving $20 in credits, valid for 3 days, to try OpenClaw before committing to paid hosting.
Which chat apps can I use to talk to OpenClaw?
It supports 29 chat channels, including WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Matrix, and IRC, so you can message it from an app you already use.
Does OpenClaw remember previous conversations?
Yes, it keeps context about you across conversations by writing to a markdown file. One commenter said this memory approach reportedly breaks down with longer-term usage, though that wasn't confirmed by the reviewer's own hands-on testing.

OpenClaw alternatives

  • Lindy

    A closed-source, cloud-only, no-code "AI employee" builder aimed at teams — the trade for OpenClaw's fully open, self-hosted, single-operator design is Lindy's managed hosting and event-triggered workflow builder.

  • n8n

    A self-hostable, open, visual workflow-automation platform built around connecting 400+ apps on a canvas, for teams that want to design multi-step automations rather than OpenClaw's message-an-agent-directly, skills-based shape.

  • Claude Agent SDK

    Anthropic's own official, code-first SDK locked to Claude models, for developers building a custom agent into their own application rather than running a ready-made, model-agnostic personal assistant.

  • Compare the whole category

    Every agent in Agent platforms, side by side on the same fields.

Anything to add?

This page is researched and re-verified by our editors — but the people who actually use OpenClaw know things we cannot see from the outside.

Report an issueSomething wrong, outdated or a dead link? Tell us and we re-verify.
Write a reviewUsed OpenClaw? Your experience helps the next buyer decide.
Get a quote from OpenClawTell us what you need and we make the introduction.
Get weekly Agent platforms updates

New agents and fresh verdicts. One short email, every week. One-click unsubscribe.