dyad

Positioned as a self-hosted alternative to cloud app builders like v0, Lovable, and Replit Agent, Dyad generates full-stack applications from natural language descriptions using locally-run LLMs or API-based models, with the generated code running entirely within the user's own environment.

The platform is designed for users who want the productivity of AI-assisted app generation with the privacy and control of local execution.

Dyad connects to local model servers (Ollama, LM Studio) as well as cloud APIs (OpenAI, Anthropic), giving users flexibility to choose their inference backend based on privacy requirements and capability needs.

Generated applications are standard codebases that users own completely no vendor lock-in, no recurring platform fees, and no restrictions on how the resulting software is used or deployed.

Dyad targets developers, researchers, and technical power users who want to rapidly prototype tools, internal dashboards, data visualization apps, or automation scripts without the overhead of full-stack development from scratch.

The local-first architecture makes it suitable for regulated industries where cloud-based AI tools are prohibited, for offline environments, and for users who work with sensitive data that cannot leave their machine.

As an open-source project, Dyad's codebase is inspectable and extensible for teams that want to adapt it to specific internal tooling workflows.