A self-hosted AI assistant that connects messaging channels, models, and tools to carry out personal or team tasks.

OpenClaw is an open-source AI assistant you run on your own computer or server. It connects chat channels to models and tools, so conversations can lead to actions rather than text alone. It supports personal use and shared team deployments through its Gateway.

Key Highlights

Messaging channels include WhatsApp, Telegram, Discord, and Slack.
Gateway manages sessions, tools, and channel connections.
Control UI and terminal interfaces provide direct access.
Hosted and local models can be configured as providers.
Skills and plugins extend the assistant's capabilities.

What Makes It Different

OpenClaw separates the assistant's workspace and channel connections from the model that powers it. You can change providers while keeping a locally managed setup. Memory and credentials live on your hardware, but prompts still reach the model providers and chat platforms you choose. Self-hosting does not make every connected service local.

Features & Capabilities

Onboarding sets up model access, a workspace, and the Gateway. Open the Control UI to try a conversation, then connect a messaging channel. Companion apps and device nodes add capabilities on supported platforms; availability depends on the platform and configuration.

Tools, skills, and plugins expand what the assistant can do. Configure access around the tasks you actually need. Main-session tools run on the host unless sandboxing is configured, so review that setup before allowing other users to interact with it.

User Ratings and Testimonials

The product website publishes community testimonials, rather than a standardized independent review score. Evaluate a small real task first: check channel reliability, tool permissions, result quality, and the amount of setup and maintenance your workflow needs.

Pricing & Value

Open-source software: Free under the MIT license, with no OpenClaw software subscription or hosted tier.
Model access and operation: Your chosen model provider, connected services, and hosting can have separate charges. Local models require suitable hardware.

The software license covers the assistant, not an unlimited allowance for paid models or external services.

Product facts

Pricing: MIT-licensed software is free to self-host; models, connected services, and infrastructure may have separate costs.^[1]
Access: Run the agent on your own computer or server and configure your model provider and messaging integrations.^[1]
Product status: An active open-source personal agent maintained through its official repository and documentation.^[1]