OpenClaw
OpenClaw is an open-source, self-hosted AI agent platform that runs directly on a user’s own hardware rather than on third-party servers. Designed as a privacy-first personal assistant, it connects to popular chat apps to automate tasks, perform system actions, and coordinate across tools. Emerging from the Clawdbot and Moltbot projects, it has become a prominent example of the “agentic” AI model that works autonomously beyond simple chat.
Key facts
Initial release: November 2025 (as Clawdbot)
Current version: v3.0 (2026)
Creator: Peter Steinberger and open-source community
License: MIT open-source
Supported systems: macOS, Linux, Windows (WSL2)
Architecture and operation
OpenClaw uses a two-part structure: a Gateway that manages sessions, authentication, and channel routing, and an Agent process that runs AI models and executes commands. It can operate entirely offline, storing all data in a local workspace directory. Users may integrate hosted models such as Claude, GPT-4, or local engines through Ollama, with conversations and configuration data kept on-device for full control .
Core capabilities
The system links directly with WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and others, turning those chat surfaces into command interfaces. It can read and send emails, manage calendars, browse the web, control files and smart-home devices, and run scheduled jobs or reminders. Persistent memory allows context to persist for weeks, enabling proactive behaviors such as generating summaries or initiating follow-ups without new prompts .
Security and privacy model
All information—including chat histories, API keys, and workflow scripts—is stored locally in encrypted form. By default the Gateway binds only to localhost to prevent remote exposure. Later releases added mandatory authentication tokens and a security --audit tool following reports of unsecured community instances. The project’s ethos centers on “local sovereignty,” meaning users retain ownership of both computation and data .
Use and customization
Installation is accomplished with a one-line shell command or Docker image. Developers extend functionality through a TypeScript-based plugin system that supports more than 50 built-in integrations, from GitHub to Notion. Advanced users can define custom “Skills” for automation or have the agent generate its own. OpenClaw’s open-source model has fostered a rapidly growing ecosystem of contributors and community-maintained extensions .
Notability
By merging conversational AI with direct system control, OpenClaw represents a shift toward autonomous, user-controlled agents. Its popularity in developer circles—topping 100 000 GitHub stars within months—reflects a wider movement toward transparent, programmable alternatives to cloud-hosted assistants.
