An agent rewrites a page overnight and you hear about it from a customer.
Nothing goes live on its own
Everything an agent writes stays a draft. It reaches the public only when a person publishes it.
Content infrastructure for AI-built and developer-built products
Change content without changing your app
Your team updates content in a dashboard, your agents update it through MCP, and your app just reads it. Routine changes stop being code changes.
Free Sandbox plan · no card required
The problem
AI can build the site. Someone still has to manage the content.
AI coding tools and developers are good at building interfaces. But once the product is live, changing a headline, publishing an article, updating a product description or adding a new page should not mean editing application source code — or asking an AI agent to rebuild the frontend again.
StepItem.tsx
type Props = { step_number?: string; title?: string; description?: string; icon?: string; [x: string]: any;}; export default function StepItem({ step_number, title, description, icon, ...props}: Props) { return ( … );}It declares the fields it needs. Nothing else.
Those same fields, filled in — and the page updating beside them.
The same four fields, in both places
step_number
title
description
icon
-
Your frontend stays yours
Build it with your preferred framework, your own developers or an AI coding tool. Garchi does not replace your frontend, your hosting or your application architecture.
-
Garchi holds the content
Pages built from reusable sections, articles and products as data items, categories, metadata, uploaded assets and per-language content all live outside your application code.
-
People and AI can manage it
Your team uses the dashboard. Your application reads the REST API. Compatible AI tools connect through Garchi's MCP server.
Product tour
See Garchi CMS in action
BUILT-IN MCP SERVER
Give AI agents structured access to your content
MCP is an open protocol that lets AI tools connect to external systems. Garchi's hosted MCP server lets a connected, compatible AI tool work with the content in your Garchi account — creating pages, updating data items, reordering sections and uploading assets against your actual content model, instead of guessing at it from your frontend source code.
CONNECT YOUR AI TOOL
MANAGE CONTENT BY CHATTING
20+ TOOLS
WORKS WITH COMPATIBLE MCP CLIENTS
YOUR AI AGENT
The agent reads your content model and calls the same Garchi operations your team uses in the dashboard, with authenticated access to the connected account.
Starter kits
Scaffold a connected project with one command
Starter kits for Next.js, Nuxt, SvelteKit and Laravel. Or skip the kit and call the REST API from any stack.
npx @lumenharbor/garchi-starter-kit -k next
The part that makes it safe
Let an agent near your content. Not near your live site.
Nobody is worried that the AI can't write. They are worried about what it does when it writes the wrong thing. Garchi is built so those mistakes stay reversible, and stay off your live site.
You spot the mistake days later, and nobody remembers what it said before.
Every version is kept
Each change becomes its own restore point, labelled with who made it. Putting the old one back takes a click.
Reading a diff doesn't tell you whether the page actually looks right.
See it on your real site first
Drafts render on your own frontend, visible to you and nobody else, until you are happy with it.
Last deploy
6 days ago
Last content change
4 min ago
Fixing one word means a pull request and a deploy.
Nobody deploys to fix a typo
Content lives outside your codebase. Whoever changes it, you ship code only when the code changes.
Also included
Multilingual content per space
Hosted, maintained and upgraded for you
A REST API any stack can call
Node.js and PHP SDKs
Notes on your content that agents can read
Separate spaces with per-space collaborators
Who it's for
Who Garchi is for
Different teams reach Garchi from the same place: the frontend is built, and the content behind it still needs managing.
Use case one
AI-built websites and applications
Use an AI coding tool to build the frontend, and keep routine business content out of the generated application code — so updating a headline or publishing an article doesn't mean another round of prompting.
Use case two
Startup and SME products
Give non-developers a place to manage landing pages, articles, product information and campaign copy, while developers keep control of the application architecture.
Use case three
Agencies and freelancers
Build the frontend for a client and hand them a dedicated place to manage ongoing content. Paid plans include multiple spaces, so separate projects stay separate.
Content modelling
Content models that fit your product
Your blog post has authors. Your product page has pricing tiers. Your case study has industry tags. Garchi's data items start with a base schema and extend with the metadata your content needs — without changing your application's database schema.