MCP server for Plyo — the home for AI-built projects. Every version saved, one click to undo, one click to publish, plus a security check before going live, visitor stats, health monitoring and database backups. Built for people who build with AI and don't code; this server is how their AI does the technical part.

Setup

Claude Code, one line:

claude mcp add plyo -e PLYO_TOKEN=plyo_xxx -- npx plyo-mcp

Any other MCP client:

{
  "mcpServers": {
    "plyo": {
      "command": "npx",
      "args": ["plyo-mcp"],
      "env": { "PLYO_TOKEN": "plyo_xxx" }
    }
  }
}

PLYO_TOKEN is an AI key from the Plyo dashboard (Settings → AI keys → "Give your AI a key"). New here? Plyo is free to start. plyo.dev has the full picture; setup lives at plyo.dev/connect.html.

After setup, the user says: "I've added Plyo. Save this project there and keep saving as we go." That's the whole integration.

Tools

Tool	What it does
`create_project`	Create a Plyo project and link the current folder to it
`link_project`	Link the current folder to an existing project by slug
`save_checkpoint`	Save the current state to the project's timeline (pass `intent`: one sentence on what the user asked for; it powers timeline search)
`publish`	Publish the latest checkpoint to the live URL; on failure returns a plain-English cause and a fix you can act on
`get_build_error`	Latest build status; if it failed, the translated error and fix guidance
`start_draft`	Safe copy of the project to try changes without touching the live version
`restore`	Return to an earlier checkpoint; never destructive, history is kept
`get_timeline`	Recent checkpoints with what changed in each
`backup_data`	Snapshot the project's database and storage files before risky data changes
`list_projects`	The user's projects with names, slugs and live URLs

Notes for agents

Call save_checkpoint after each completed piece of work, with a one-sentence intent.
Call backup_data before schema migrations, bulk updates or deletions.
Errors come back plain and actionable: publish and build failures include what went wrong and what to change.
Under the hood every project is a real git repository; Plyo also accepts plain git push at https://git.plyo.dev/{slug}.git (username plyo, the AI key as password).
A failed publish never takes a live site down.