Content that ships at the speed of AI.
A code-first, type-safe headless CMS with a real admin and native AI. Define it in TypeScript, run it on any framework, and own every layer.
In private beta. Free and open source, MIT licensed, works with any framework.
// Your content model IS TypeScript. Reviewed, versioned, pushed. import { defineObjectType, field } from '@tangen/cms' export default defineObjectType({ name: 'post', localized: true, fields: { title: field.text({ required: true }), body: field.richtext({ localized: true }), author: field.relation({ targetType: 'author' }), }, })
One content core powers every surface you build. Websites, mobile apps, native iOS and Android, even IoT, all over REST, GraphQL, or a typed SDK.
- Next.js
- Nuxt
- Svelte
- Astro
- React Native
- Flutter
- iOS
- Android
A backend that reads like code, and reasons like an agent.
AI-native, not bolted on
MCP and RAG ship in the box. Agents read, write, and reason over your content through typed, permission-scoped tools.
Schemas as TypeScript
Define types with defineObjectType. Your model lives in git, reviewed in pull requests, versioned like code.
Typed end to end
Types flow from schema to SDK. A wrong field breaks the build at your desk, not production.
Workspaces and environments
Many workspaces, one install. Dev, staging, and production, data-isolated. Push per environment, promote when ready.
Live preview, included
Editors see drafts render on the real site as they type. AI edits preview too, from the same state.
Batteries included
Roles, localization, flows, jobs, webhooks, media, forms, and a full admin. All optional, never in your way.
Fast by default, from install to request.
Tangen runs on Bun and Elysia, a modern runtime and a Bun-native web framework, both built for speed. You feel it in the small moments that add up: installs, cold starts, and requests that come back before you blink.
- Bun
- Elysia
- Postgres
- Drizzle
- TypeScript
- React
- 30x
- faster installs
- vs npm
- 4x
- faster startup
- vs Node.js
- 0
- build step
- native TypeScript
- 1
- toolchain
- runtime to tests
Figures are from Bun's published benchmarks. Real numbers depend on your project and machine.
A real admin, not a config screen.
This is what your team works in every day, not a wall of settings.
See the page as you write it.
Edit content and watch it render on your real site in real time. No publish-and-refresh loop, and no premium add-on to unlock it.
Model content without leaving the browser.
A visual builder that writes real TypeScript back to your repo, so designers and developers stay in sync on one schema.
Describe it. Watch it build.
Ask for a type, a draft, or an image and the assistant writes into the same draft your editors see, previewing live as it goes.
Two kinds of developer. One backend both love.
Whether you build by talking to an agent or by writing every line yourself, Tangen meets you where you are, and never makes you choose.
Talk to your project. Watch it build.
Describe to build. Explain a feature in plain English and get real schema and drafts you confirm.
Content AI. Generate and fill fields, rewrite copy, and draft entries, all inside the editor.
Bring your agent. Point Claude Code or Cursor at it over MCP and let it model, seed, and wire content.
Instant feedback. AI edits stream straight into live preview, so you watch the result as it happens.
Own every layer, typed from schema to query.
Typed SDK. A fully typed client where a wrong field is a compile error, not a runtime surprise.
TangenQL. A clean, expressive query language for filtering, sorting, and relations across REST and GraphQL.
API playground. Explore and test the REST and GraphQL API right in the admin, with real auth.
Code-first, in git. Schemas, routes, and config are TypeScript in your repo. Review them, version them, ship them.
Building for editors, not developers? See Tangen for marketers
Your content, operable by AI.
Most CMSs bolt AI on as a chat box. Tangen makes your content a first-class surface for agents, safely. Point Claude Code, Cursor, or any MCP client at your project and let it build with you.
Agent tools, permission-scoped
Agents call list_types, search, create, and publish through the Model Context Protocol. Every call is bounded by the key's grants, and every write is attributable.
RAG over your content
Semantic search and grounded answers across your content, docs, and tables. pgvector under the hood, real citations on top.
Describe it, and it builds
A schema and content assistant turns plain English into real types and drafts you confirm. It proposes, you approve. Nothing is applied blindly.
# point any MCP client at your running project $ claude mcp add tangen -- tangen mcp --key <api-key> # the agent now has typed, grant-scoped tools: > list_types the content model > search_content query with TangenQL > create_object writes land as drafts > ask_docs RAG, grounded and cited # publishing stays a human decision.
You decide what the AI can see, touch, and sound like.
AI that can write to your content is only safe if you set the boundaries. Tangen keeps them in your hands, in code.
Scoped by role
Agents get exactly the grants their API key allows. Every write is attributed, and no key means no access.
Field-level control
Hide any field from AI reads and writes with a single flag, so sensitive data never reaches a model.
Your content style
Define the voice, tone, and rules the assistant writes in, so drafts sound like you, not a chatbot.
Human in the loop
AI proposes, you approve. Writes land as drafts, and publishing stays a human decision.
Familiar tools, taken further.
Tangen borrows the best ideas from the tools you already know, adds an AI-native core, and then gets out of your way on pricing.
A database, plus the CMS.
Supabase nails the database and APIs, then leaves the content layer to you. Tangen ships that layer: typed content, a real admin, localization, and AI, all ready to use.
See the full comparisonvs Payload & code-first CMSCode-first, and AI-native.
Like Payload and Strapi, Tangen is code-first, TypeScript, and self-hosted. It adds a framework-agnostic admin and native AI, with MCP and RAG in the box, not a plugin.
See the full comparisonvs hosted SaaS CMSYours, with no paywalls.
Contentful, Sanity, and Contentstack rent you a platform. Tangen you own outright: no per-seat pricing, no metered calls, and no paywall on preview, roles, or environments.
See the full comparisonComparisons reflect the common positioning of each tool and are meant to help you choose, not to knock anyone. Every product named here is excellent at what it set out to do.
Define. Push. Build. Deploy.
From npm install to a running blog backend, with a typed API and an admin, in about ten minutes.
Blog backend in ~10 minutes- 01
Define
Model your content in TypeScript. Types, fields, relations, and localization, all living in your repo.
- 02
Push
Run tangen push to migrate your database and generate types. Idempotent, reviewable, and per environment.
- 03
Build
Query over REST, GraphQL, or the typed SDK from any framework. Then hand the same project to your AI.
- 04
Deploy
Self-host anywhere you already run things. Any cloud, any region, your infrastructure, your data.
$ tangen create my-app $ tangen push # migrate and generate types // then, from your app, fully typed: const { data } = await tangen .objects('post') .where({ status: 'published' }) .sort('publishedAt', 'desc') .get()
Build on Tangen in the private beta.
Define a content type, push it, and query it from anywhere, with your own AI agent doing the rest. Request access and we will get you set up.