Skip to content
Andesine

Local server and plugin

Run the Andesine MCP server locally for coding agents, and install the Claude Code and Cursor plugin.

Coding agents, such as Claude Code, Cursor, Codex, and Gemini CLI, can run the Andesine MCP server on your computer with the CLI. The local server has the same tools as the remote server, and one more: upload_image, which uploads image files from your computer.

Set up a project

  1. Sign in

    Run npx andesine auth login. The CLI opens Andesine in your browser. Enter the code from your terminal on the Authorize CLI page and select Approve. In CI or other environments without a browser, set ANDESINE_API_KEY instead.

  2. Add the server to the project

    In your project, run npx andesine init --mcp. Choose the workspace. The command saves it in andesine.json and adds the server to the files of the agents that the project uses.

  3. Start your agent

    Start the agent in the project folder. It starts the server with npx -y andesine mcp --workspace <id>.

init --mcp writes to these files, and keeps the other servers and settings in them:

File

Agents

.mcp.json

Claude Code, VS Code

.cursor/mcp.json

Cursor

.gemini/settings.json

Gemini CLI

.codex/config.toml

Codex, in trusted projects

You can also add the server with the command of your agent:

sh
sh

Options

  • Servers added with init --mcp use the workspace in their --workspace option. Other servers, such as the plugin's server, use the workspace in the nearest andesine.json.

  • --read-only lists only the tools that read content.

  • With an API key, the server lists only the tools that the key's permissions allow.

Claude Code and Cursor plugin

The Andesine plugin adds the local server and a skill that tells the agent when to use the MCP tools, the CLI, or the SDK.

In Claude Code, run:

text

This needs Claude Code 2.1.275 or later.

In Cursor, install Andesine from the Cursor Marketplace.

Before you use the plugin, sign in with npx andesine auth login and run npx andesine init --mcp in the project.

Privacy

The local server runs on your computer and sends requests only to Andesine, with your CLI sign-in or API key. It sends the content that the agent reads or writes and the images that it uploads.

Last updated