AI-native content engine
Content modeling, persistence, media pipelines, AI orchestration, and static output — all in one embeddable TypeScript library.
Provider-agnostic agents for content generation, rewriting, translation, SEO optimization, and design token generation. Swap between Anthropic, OpenAI, or local models.
@webhouse/cms-aiJSON Schema powered collections and blocks. Every piece of content is typed, validated, and introspectable — so AI agents can reason about structure.
cms.config.tsThe production artifact is always pre-rendered HTML + CSS + minimal JS. No runtime framework unless you opt in. Incremental builds with dependency tracking.
npx cms buildFirst-class integration with Next.js (App Router, Server Components, ISR), Astro (island architecture), and generic Node.js/Express.
@webhouse/cms-adapter-nextSharp-based image processing with AI generation (Flux, DALL-E), responsive variants, WebP/AVIF conversion, blur hashes, and auto alt-text generation.
@webhouse/cms-mediaField-level protection enforced at the engine level. AI agents can never unlock fields — only humans can. WriteContext actor threading through all CRUD operations.
PATCH-AI-LOCKA pipeline of discrete stages. Each can be extended, replaced, or bypassed.
Two MCP servers. One public and read-only — bundled with every built site. One authenticated — for content production from Claude, Cursor, or any MCP client.
Bundled with every site. Any AI agent can discover and query published content — no API keys, no documentation needed.
Full content production from Claude iOS, Claude.ai, Cursor, or Claude Code. Create, edit, publish, generate with AI — all via MCP.
Then keep iterating with AI or by hand.