Introducing Mosaic Agent Skills
AI-Powered Code Generation for Mosaic Development
Building a complete Mosaic station means getting a lot of things right: GraphQL operations, TypeScript components, form validation, route registration, breadcrumbs, integration with existing stations. The architecture is solid — but following every convention correctly is hours of careful, repetitive work.
Mosaic Agent Skills is an MCP server (@axinom/mosaic-agent-skills) that works with Claude Code, GitHub Copilot, and any MCP-enabled AI agent. Write a prompt:
The agent discovers your project structure, reads your GraphQL schema, and returns a complete, production-ready implementation in minutes — not a skeleton, but fully wired code following Mosaic best practices automatically.
What Gets Generated
The initial release covers the three core station types:
-
Explorer Stations — entity browsing with filters, sorting, pagination, real-time updates, bulk-edit, and quick-edit
-
Create Stations — entity creation forms with validation
-
Details Stations — full editing interfaces with update/delete actions, image and video management, and resource management
Every generated station includes its GraphQL operations, TypeScript components, validation, route registration, and breadcrumbs — ready to run.
What’s Coming Next
Frontend stations are the starting point. Backend generation, brownfield updates to existing code, and extensibility tools so teams can build their own project-specific skills are all on the roadmap.
The full blog post covers setup, real usage examples, and the complete roadmap.
