AI Agents Run Our Website: Content Management through MCP
We use our own technology first. The news on this website are written, reviewed and published by AI agents through an Admin MCP server — the same Model Context Protocol approach we build for our clients.
- Client
- Asapeca (internal project)
- Scope
- News publishing and media uploads
- Standard
- Model Context Protocol (MCP)
Technologies
- Model Context Protocol
- TypeScript
- SvelteKit
- Markdown
The Challenge
A company website needs regular updates, but a traditional CMS brings a back office to host, secure and learn — overhead that a small engineering team would rather avoid.
At the same time publishing must stay safe: nothing goes live without a human seeing it first.
The Solution
Admin MCP server
Content tasks are exposed as MCP tools, so any MCP-capable AI agent can manage the news section in natural language.
Draft-to-live flow
The agent creates a draft and gets a private preview link; the post is published only after a person has reviewed and approved it.
Media uploads
Images are uploaded through the website’s admin API and referenced in the post.
Content as files
Posts are Markdown files on the server and appear on the site immediately, without a rebuild or a redeploy.
Key Results
Instant updates
A news post goes from idea to published page in one conversation with an agent.
Human in the loop
Every post passes a preview and an explicit approval before it goes live.
No CMS to maintain
Access is secured with SSH and API keys; there is no admin UI to host or patch.
Have a similar challenge?
Tell us about your systems and processes — we will show how an integration or an AI agent can take the manual work off your team.