Start with MCP, CLI, SDK, rules, and skills
Developers get a local-first content stack: CLI setup, MCP tools, rules, skills, validation, generated SDK access, and a clear path into Studio when teams need review.
CLI
Initialize a content system inside the repository
The CLI creates a .contentrain workspace where models, content, metadata, vocabulary, and context live beside the application. That gives AI coding agents and developers a source of truth they can read before they change copy.
- contentrain init, status, doctor, and validate for project setup
- Model kinds for singletons, collections, documents, and dictionaries
- Plain JSON and Markdown instead of hidden CMS state
MCP
Use MCP as the agent operating layer
The MCP package exposes bounded tools for content, models, validation, scanning, normalize, branch submission, and bulk operations. Agents can call the same capabilities repeatedly instead of inventing repository-specific write logic.
- Local stdio transport for desktop agents
- HTTP transport for remote or embedded agent workflows
- Provider interfaces for local, GitHub, and GitLab execution paths
Normalize
Normalize hardcoded strings before they become debt
The normalize workflow scans source files, extracts approved strings into content models, then patches code to reuse generated content access. It turns the most common AI-built app problem into a reviewable workflow.
- Scan candidates and source graphs
- Extract to Contentrain content without touching source first
- Reuse generated content access after review
SDK
Generate runtime access for modern frameworks
The JavaScript SDK generates query clients that Nuxt, Next, Astro, SvelteKit, Vite, Expo, and Node apps can consume without a CMS runtime dependency. The app imports content through stable typed APIs instead of parsing files by hand.
- Generated #contentrain style imports
- Locale-aware query access
- Works with static and server-rendered builds
Rules and skills
Ship shared rules and skills with the content system
The rules and skills packages make content operations legible to agents. They define how to validate schemas, normalize strings, manage translations, review content, and avoid unsafe edits.
- Reusable workflow skills for init, content, model, validate, normalize, translate, and review
- Quality rules for SEO, i18n, accessibility, schema, media, and security
- Context bridge that explains the project to the agent
Studio path
Move to Studio when collaboration becomes the bottleneck
The developer path is intentionally not a dead end. When non-developers need review, roles, media, CDN delivery, forms, or usage controls, the same content contract can be operated through Studio.
- Start local with MIT packages
- Keep Git as the content audit layer
- Add Studio for team operations and delivery
Common questions
Do developers need Studio to start?
No. The MIT packages let you initialize, validate, normalize, generate, and serve local workflows before a Studio account exists.
What is the first command for a new app?
npx contentrain init is the default entry point. From there, run validate, generate a query client, and use normalize when the app already has copy in source files.
Why is MCP important here?
MCP gives AI coding agents a bounded, repeatable interface for content operations. That is safer than asking an agent to edit arbitrary JSON and source files from scratch.
Next paths
Continue through the content system
Developer implementation playbook
Install, model, validate, generate, and ship runtime access.
AI agent governance playbook
Bound MCP tools, rules, review, and validation for coding agents.
Normalize migration playbook
Move hardcoded UI copy into structured content safely.
Integrations
See agent, framework, provider, and deployment paths.
Start local. Scale to Studio.
Build a governed content layer before content becomes product debt.
Developers can start with the MIT packages. Teams can move into Studio when review, roles, delivery, and licensing matter.