Dyad, the free and open-source AI app builder that runs locally on users' machines, is gaining traction in the developer community as an alternative to cloud-based platforms like Bolt and Replit. However, recent community discussions reveal both promising developments and significant technical challenges that users are encountering.
Platform Availability:
- macOS (Apple Silicon and Intel)
- Windows
- Linux support in development
- Downloaded 100,000+ times with 4.9/5 user reviews
Token Consumption Issues Limit Large Project Development
One of the most pressing concerns raised by the community centers around excessive token usage when working with larger codebases. Users report that AI-powered app builders like Dyad consume tokens at an alarming rate, particularly when projects exceed a few thousand lines of code. This limitation effectively restricts the platform's usefulness for complex applications.
These things does work (up to a max few thousand lines of code), but, they eat tokens like f*ng monsters.
To address this challenge, Dyad has implemented manual context management features that allow developers to selectively choose which files to include in the AI's context window. The development team is also exploring more sophisticated solutions, including code embeddings and potential integration with specialized tools that could help optimize token usage through intelligent code analysis.
Key Technical Features:
- Local execution for privacy and speed
- Support for multiple AI models (Gemini 2.5 Pro, GPT-4.1, Claude Sonnet 4)
- Ollama integration for local model execution
- Supabase integration (Auth, Database, Server Functions)
- Manual context management for large applications
Security Concerns Around Database Access Controls
Community members have raised important questions about data security, particularly regarding Supabase integration and row-level security (RLS) implementation. Some users worry about potential security vulnerabilities where improperly configured database access controls could expose user data across different accounts.
The Dyad team has responded by implementing extensive system prompts designed to guide the AI in following RLS best practices. However, they acknowledge that some security behaviors stem from the underlying Supabase platform itself rather than the app builder, highlighting the complexity of maintaining security in AI-generated applications.
Platform Limitations and Integration Challenges
While Dyad excels at creating applications using modern JavaScript frameworks, some users have encountered limitations when working with older technologies. Bootstrap-based websites, for example, present challenges since they don't align with the platform's focus on JavaScript frameworks. This restriction may limit adoption among developers working with legacy systems or specific design frameworks.
Dyad Pricing Plans:
- Dyad Free: $0 USD/month - Local app builder, no sign-up required, community support
- Dyad Pro: $30 USD/month - Exclusive AI Pro modes, 300 AI credits/month, direct support
- Dyad Max: $79 USD/month - 900 AI credits/month, reload credits anytime
Future Developments and Community Engagement
The development team is actively working on expanding Dyad's capabilities, with plans to add Model Context Protocol (MCP) support that could integrate with specialized tools for better code management and analysis. The community is contributing recommendations for useful MCP servers, including Context7 and Deepwiki for enhanced information access.
Creator Will Chen continues to emphasize Dyad's core philosophy of providing developers with true ownership of their tools and code, maintaining the platform's commitment to remaining free, open-source, and locally-run. This approach resonates with developers seeking alternatives to subscription-based cloud platforms, even as technical challenges remain to be solved.
The platform's growing user base and active community engagement suggest strong potential, but addressing token efficiency and security concerns will be crucial for broader enterprise adoption.
Reference: Build apps with AI, no coding required