Let your assistant make a server.

The included MCP server exposes the same project operations as the web editor over stdio, so an AI assistant can write, save, and deploy drops.

Connect an MCP client

The MCP server is a small Node.js script in the jsdrop source (scripts/mcp.mjs). Add this entry to your MCP client, with the path to that script. No API key is needed for public guest drops.

{
  "mcpServers": {
    "jsdrop": {
      "command": "node",
      "args": ["/absolute/path/to/jsdrop/scripts/mcp.mjs"],
      "env": { "JSDROP_URL": "https://jsdrop.com" }
    }
  }
}

Available tools

• create_project, get_project, update_project, fork_project
• set_index_ts, set_package_json
• create_version, get_version, list_versions, restore_version
• deploy_project (the drop’s owner, with Pro)

Try a prompt

Create a jsdrop with a Hono API that converts Celsius to Fahrenheit.
Save a version and give me its public URL.

How edits work

Tools edit the working state directly. Changes arrive in connected browser editors through Yjs. The server checks revisions before applying state, and versions remain immutable. Save a version after editing to update the public page and any deployment. Deleting drops, private drops, and account operations are not available through MCP.

Building an MCP server instead?

To give an assistant new tools, start from the MCP server starter kit: a Model Context Protocol server over HTTP, in one index.ts. Deploy it with Pro for an endpoint that is always on.