Open source · Local-first deployment · Cross-platform access
OpenClaw
AI agents built for execution, control, and professional workflows
OpenClaw is a local-first, open-source AI agent platform designed for real operational work. It can connect to chat platforms, manage calendars and email, assist with code, browse the web, process files, and run recurring tasks inside clearly defined permission boundaries.
🪟
Windows
🍎
macOS
🐧
Linux
🤖
Android
📱
iOS
☁️
Cloud
Move the pointer over the agents to see the interaction animation.
⟋⟋ Download
Choose the deployment path that fits your environment
OpenClaw supports Windows, macOS, Linux, Android, and iOS. The same platform can be used across desktop, mobile, and cloud access paths while preserving a more consistent operating model and data boundary.
Windows
macOS
Linux
Mobile
Email automation
Classify inbox traffic, summarize messages, draft replies, and reduce turnaround time across routine communication.
Calendar & scheduling
Coordinate calendars across platforms with conflict detection, reminders, and scheduling logic built for team workflows.
Code assistant
Support review, bug fixing, test generation, pull-request workflows, and remote CI/CD actions inside engineering processes.
Automation workflows
Handle browser actions, form submission, extraction, and multi-step orchestration through reusable execution patterns.
Sheets & analytics
Import structured data, generate charts, and surface insights that can be used directly in reporting and decision support.
Smart home
Control connected devices through natural language and unify home automation actions into a single interaction layer.
⟩ Unified access to leading models
Reduce integration overhead while keeping model choice flexible
Connect major providers through one interface, simplify maintenance, and adapt routing without rewriting business logic.
Mainstream models & providers
OpenAI · GPT-4o / o1
Claude · Opus / Sonnet
Gemini · Pro / Ultra
DeepSeek · V3 / R1
Qwen · Qwen Max
Doubao · Doubao Pro
Moonshot · Kimi
Zhipu · GLM-4
Baidu ERNIE · ERNIE 4.0
Minimax · abab6.5
A unified layer makes it easier to balance quality, latency, availability, and cost, while preserving the freedom to switch providers when requirements change.
Unified API
Use one connection pattern across providers and reduce repetitive integration and maintenance work.
Cost-aware routing
Match tasks to the right model path so budget is spent where it changes outcomes most.
Fast model switching
Compare outputs, latency, and quality quickly when evaluating the best setup for a workflow.
⟩ Visual control
Manage deployment and policy through a graphical control layer
Monitor runtime status, configure providers, connect platforms, and adjust permission rules without relying on command-line operations for every task.
Check runtime status in real time and perform common control actions directly from the dashboard.
Configure multiple providers, endpoints, and access settings through one consolidated interface.
Bring agents into existing chat environments and manage connection status from a single view.
Click any screenshot to open a larger preview.
⟩ Integrations
Bring AI agents into the communication tools your team already trusts
Issue instructions, review results, and keep execution inside established collaboration channels instead of forcing teams into an unfamiliar interface.
WeChat
Feishu
DingTalk
WeCom
Telegram
Discord
Slack
Signal
⟩ Capabilities
From conversational interface to executable work
OpenClaw can read files, run scripts, operate a browser, and coordinate tools inside the permission model you define.
Persistent memory
Retains preferences, operating patterns, and relevant context so collaboration becomes more consistent over time.
System-level execution
Read and write files, execute scripts, and operate browser tasks with sandboxed or expanded permissions defined by policy.
Multi-model compatibility
Work with Claude, GPT, DeepSeek, Qwen, and local open-source models inside one operational framework.
Local-first control
Keep data, configuration, and runtime context closer to your own environment for stronger oversight and privacy control.
Proactive operations
Run scheduled tasks, monitor background conditions, and surface alerts before routine issues become bottlenecks.
Extensible behavior
Expand skills, refine execution logic, and adapt workflows to fit changing operational requirements.
⟩ Governance and control
Action approval and workspace boundaries keep execution accountable
Every material action can be authorized, limited, and reviewed so execution remains useful without becoming opaque or uncontrolled.
Command-level approval
Use allowlists, denylists, and approval boundaries to define which commands can run and under what scope.
Sandboxed file access
Restrict readable and writable paths to specific workspaces so agents cannot move beyond approved storage boundaries.
⟩ Quickstart
Initialize the environment and connect a workspace in minutes
A common setup path includes installation, onboarding, platform connection, and the start of task execution inside an approved operating context.
openclaw onboard openclaw connect --platform wechat
Once connected, tasks can be issued through familiar communication channels while context, approvals, and results remain visible.
Typical task coverage
UI demo
Email
Auto-reply
Calendar
Smart scheduling
Web
Form filling
Dev
Fix tests
Approvals and restrictions remain in place so execution can move quickly without losing oversight.
⟩ What users say
Feedback from early users and developers
These reactions highlight execution quality, local control, and the practical value of agents that can do more than generate text.
“After years of AI hype, I thought nothing would impress me—until I installed @openclaw.”
@lycfyi
“It’s like a coworker with eyes and hands at your computer—message it and it can do anything you can.”
@nathanclark_
“Persistent memory, persona guidance, chat integrations… a few rough edges, but the experience is fantastic.”
@AryehDubois
“It’s open source—your context and skills stay on your own machine, not in a walled garden.”
@danpeguine
“OpenClaw gets a lot of the hard parts right. I’m genuinely impressed.”
@AryehDubois
“Even if models stop getting stronger, we’ll spend years finding new uses—@openclaw is that kind of leap.”
@markjaquith
“Incredible value: runs tests, catches errors, fixes them, and opens a PR. The future is here.”
@nateliason
“I ran OpenClaw for a week—you can control it and customize it instead of relying on a tech giant.”
@snopoke
“Installed @openclaw yesterday… chatting in Discord and it keeps building itself. Wild.”
@steipete
“It runs my company.”
@therno
⟩ FAQ
Deployment, permissions, and model access
Core questions about installation, governance, provider strategy, and the role of this template.
The desktop app is the faster path to operational use: visual installation, upgrade flow, configuration UI, common integrations, and safer defaults for non-developers. The open-source core is the extensible runtime layer: plugins, skills, orchestration, memory, model abstractions, and deeper customization. One is optimized for operational speed. The other is optimized for control and extensibility.
Yes. OpenClaw can operate within layered controls such as command approvals, allowlists and denylists, directory sandboxing, confirmation steps for side-effect actions, and stronger isolation through separate accounts, containers, virtual machines, or outbound network policies.
Yes. Models and providers can be managed as one capability pool. Teams can route by task type, latency, quality, or cost, centralize key and quota management, and define failover behavior when a provider is unavailable or rate-limited.
Yes, this is the official website of openclaw. It provides the latest updates, official downloads, and authorized resources. Please be cautious of third-party sites, as they may not offer legitimate products or services.
Ready to put OpenClaw into production workflows?
Move from evaluation to deployment with local-first control, extensible architecture, and execution paths designed for real operational work.