A home for every project
Keep related conversations together. Add project instructions and reference files to share context across chats in that project.
A familiar place to chat, organize projects, and work with your files—with the AI running right on your PC.
Windows 11 · Local inference · No API charges for local chats
Illustrative conversation · Your chats stay on your PC
LocalClaw brings the pieces together in a Windows desktop app, so you can focus on the conversation.
Keep related conversations together. Add project instructions and reference files to share context across chats in that project.
Guided setup checks your PC, helps you choose a model, and tests an actual response, tool calling, and GPU memory placement.
Add text, code, CSV, JSON, or Word documents. Relevant source excerpts are selected locally to accompany your questions.
Manual setup options offer another route when detection fails. Export diagnostic logs for troubleshooting, or back up your projects and chats.
Once your model and components are installed, local chat can work without an internet connection. Turn online tools on whenever you want them.
Paid services are a separate choice. They require deliberate opt-in and warnings. Paid requests display a prominent blinking charge warning. Internet access alone never enables a paid fallback.
Allow free web search and page retrieval while the model runs locally. Online tools can send searches and requests to external services.
Disable online tools and apply Windows Firewall protection to the app's private AI engines. Keep chatting with your installed local model.
The switch illustrations above describe controls inside the desktop app.
Download the Windows app and launch it. The setup wizard explains each stage as you go.
Check your PC, choose a model and download location, then let setup install the required AI components.
Run the built-in test. Once it passes, create a project, add any sources you need, and send your first message.
No. Local conversations use a model on your computer through Ollama. Optional paid providers use their own API billing and must be configured and explicitly enabled separately.
No. LocalClaw offers a familiar chat-and-project workspace using local models through Ollama and OpenClaw. Response quality and speed depend on the model and your computer. LocalClaw is not affiliated with Anthropic.
Chats and project sources are saved on your PC. Use Help / Backups in the app to export them. Optional online tools contact external services; explicitly selected paid requests send their included conversation context to the chosen provider.
Yes, after installing the required components and model. Turn Internet access OFF in the app for offline use. Web search and page retrieval need a connection.
Press Enter to send, or click Send. Use Shift+Enter to add a new line. You can stop a reply while it's generating.
Local models. Organized projects. Clear controls.
Start with guided setup and make it your own.
Windows 11 desktop app · v0.1.3
A question, idea or bug report for LocalClaw. Up to 500 characters; 10 messages per rolling hour.