Skip to main content

Connect ARES Commander to AI Assistants with the ARES MCP Server (2027 SP2)

ARES Commander 2027 SP2, MCP, Model Context Protocol, MCP Server, AI, Claude, Claude Code, Claude Desktop, Windsurf, Codex, Gemini, MCPSTART, MCPSTOP, MCPSTATUS

Written by Christian Kramer

With ARES Commander 2027 Service Pack 2, you can connect ARES Commander to your favorite AI assistant, such as Claude Desktop app, Claude Code, Devin Desktop, Codex or Google Antigravity. This is done with the new ARES MCP Server. Once connected, you simply tell the AI assistant in plain language what you want, and it works directly in ARES Commander for you.

What is MCP?

MCP stands for Model Context Protocol. It is an open standard that lets AI assistants talk to other programs in the same way.

Think of MCP like a USB plug for AI: instead of a special connection for every AI product, there is one standard connector. ARES Commander offers this connector, and any AI assistant that supports MCP can use it.

What can I do with it?

The AI assistant can read your drawing, run commands and check the result. For example:

  • Draw from a description: "Draw a 100 x 50 rectangle with a circle in the center."

  • Ask questions about a drawing: "Which layers are frozen?", "How many blocks are in this drawing?", "List all text in the title block."

  • Check a drawing before you send it: find parts without dimensions, overlapping annotations or text that is too small.

  • Do repetitive work in one go: "Replace 'Rev A' with 'Rev B' in all text and move every circle to layer HOLES."

  • Save and share: "Save this as plate_revB.dwg and give me a picture of it."

Good to know

  • It runs on your computer: The connection between the AI assistant and ARES Commander stays on your own computer (localhost). You do not need to install extra software such as Python.

  • It works with many AI assistants: The same ARES MCP Server works with any AI assistant that supports MCP.

  • It is the same technology as the ARES Assistant: The built-in ARES Assistant palette uses the same tools. So the assistant inside ARES and external AI assistants can do the same things.

  • Windows, macOS and Linux are supported.

What you need

  • ARES Commander 2027 SP2 or newer

  • At least one AI assistant that supports MCP, for example Claude Desktop app, Claude Code, Devin Desktop, Codex or Google Antigravity.

Step 1: Start the MCP server in ARES Commander

  1. Open ARES Commander.

  2. In the command line, type MCPSTART and press Enter.

  3. When asked for a port, press Enter to use the default port 19400.

Note: You need to run MCPSTART once each time you start ARES Commander. If ARES Commander is started automatically by your AI assistant, the server starts on its own.

These three commands help you control the server:

Command

What it does

MCPSTART

Starts the MCP server. Asks for a port (default 19400). Tells you if the server is already running.

MCPSTOP

Stops the MCP server.

MCPSTATUS

Shows if the server is running, and its port, running time, number of requests and number of connections.

Step 2: Connect your AI assistant

Each AI assistant has its own settings file. In all cases you add one entry called ares that points to this address:

http://localhost:19400/mcp

Important: If your settings file already has content, do not delete it. Only add the ares entry inside the existing mcpServers section. In the paths below, replace <your-user-name> with your Windows user name.

Claude Desktop

  1. Open Claude Desktop.

  2. Go to Settings → Developer.

  3. Click Edit Config to open the configuration file.

  4. Add this entry (if the file is empty, paste all of it):

    {
    "mcpServers": {
    "ares": {
    "command": "npx",
    "args": ["-y", "mcp-remote", "http://localhost:19400/mcp"]
    }
    }
    }
  5. Save the file.

  6. Close Claude Desktop completely. On Windows, open the Task Manager and end any Claude processes that are still running.

  7. Start Claude Desktop again and check under Settings → Developer that ares is listed and running.

Claude Code

  1. Open a terminal.

  2. Run this command:

    claude mcp add --scope user --transport sse ares http://localhost:19400/mcp
  3. Start a new Claude Code session.

  4. Type /mcp to check that ares is listed and connected.

Devin Desktop

  1. Open the file mcp_config.json in this folder (create it if it does not exist):

    • Windows: C:\Users\<your-user-name>\.codeium\windsurf\

    • macOS and Linux: ~/.codeium/windsurf/

  2. Add this entry (if the file is empty, paste all of it):

    {
    "mcpServers": {
    "ares": {
    "type": "sse",
    "url": "http://localhost:19400/mcp"
    }
    }
    }
  3. Save the file and restart Devin Desktop.

  4. Open a new Cascade chat with Ctrl + L and check in the MCP menu that ares is active.

Codex

  1. Open the file config.toml in this folder (create it if it does not exist):

    • Windows: C:\Users\<your-user-name>\.codex\

    • macOS and Linux: ~/.codex/

  2. Add these lines:

    [mcp_servers.ares]
    url = "http://localhost:19400/mcp"
    enabled = true
  3. Save the file and restart Codex (or open a new session).

Google Antigravity

  1. Open the file settings.json in this folder:

    • Windows: C:\Users\<your-user-name>\.gemini\

    • macOS and Linux: ~/.gemini/

  2. Add this entry inside the file:

    "mcpServers": {
    "ares": {
    "type": "sse",
    "url": "http://localhost:19400/mcp"
    }
    }
  3. Save the file, restart Google Antigravity and open a new session.

  4. Check that ares is listed as an active MCP server.

Step 3: Test the connection

  1. Make sure ARES Commander is open and MCPSTART has been run.

  2. Open a new chat in your AI assistant and make sure the ares MCP server is enabled.

  3. Type this prompt: Draw a circle at 200,200 with radius 50

If a circle appears in ARES Commander, everything works. 🎉

Example prompts to try

  • "Draw a 200 x 120 plate with a 10 mm hole in each corner, 15 mm from the edges, and zoom to fit."

  • "Give me an overview of this drawing: which layers are used, how many blocks, and are any references missing?"

  • "Check this drawing before I send it out: is everything dimensioned and is all text readable?"

  • "Replace 'Rev A' with 'Rev B' in all text, and move every circle to layer HOLES."

  • "Save this as plate_revB.dwg in the same folder and give me a picture of it."

What the AI assistant can do in ARES Commander

Area

What the assistant can do

Session

Check that ARES Commander is ready and get version information.

Commands

Run any ARES command, LISP expressions and macros. List the available commands.

Read the drawing

Get a summary, layers, blocks, external references, entities and their properties, dimensions, text, the current selection and the undo history.

Change the drawing

Change entity properties, edit text, create multiline text and select entities.

Drawings

List open drawings, switch between them, save under a new name and check the saved file.

Quality checks

Validate the drawing, find parts without dimensions, and find overlapping or too small annotations.

View

Zoom, read view and viewport information, and save the view as a PNG image.

System variables

Read and change system variables.

User interface

Find and press buttons, fill in fields, use shortcuts and context menus, and take screenshots.

Built-in safety

  • You stay in control when saving: ARES Commander asks you to confirm before the AI assistant saves a drawing under a new name.

  • No interruptions: The AI assistant cannot start a new command while another command is still running.

Troubleshooting

  • The AI assistant cannot connect: Type MCPSTATUS in ARES Commander. If the server is not running, type MCPSTART.

  • The ares server does not show up in the AI assistant: Check the settings file for typing errors (missing commas or brackets), save it and restart the AI assistant completely.

  • Other MCP servers stopped working: You may have overwritten your settings file. Add your other servers back next to the ares entry.

  • Port already in use: Run MCPSTART with a different port, and use the same port in the address in your AI assistant settings.

Did this answer your question?