IvoryOS MCP

IvoryOS MCP server

Serve as a robot control interface using IvoryOS and Model Context Protocol (MCP) to design, manage workflows, and interact with the current hardware/software execution layer.

🧩 One-click install in Claude Desktop (desktop extension)

1. Download ivoryos-mcp-<version>.mcpb.
2. Double-click it (or drag it into Claude Desktop, or Settings → Extensions → Advanced settings → Install Extension…).
3. Fill in your ivoryOS URL, username and password, then enable the extension.

Claude Desktop sets up Python and the dependencies itself; nothing else to install. The extension runs on your computer, so it only needs to reach ivoryOS on your network.

To build the .mcpb (from this repo, needs Node.js):

npx @anthropic-ai/mcpb pack . ivoryos-mcp-<version>.mcpb

Keep version in manifest.json in sync with pyproject.toml. .mcpbignore keeps .env and build files out of the bundle.

🚀 Quickstart with Claude Desktop (manual config)

Install uv. Open up the configuration file, and add IvoryOS MCP config.

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "IvoryOS MCP": {
      "command": "uvx",
      "args": [
        "ivoryos-mcp"
      ],
      "env": {
        "IVORYOS_URL": "http://127.0.0.1:8000/ivoryos",
        "IVORYOS_USERNAME": "<IVORYOS_USERNAME>",
        "IVORYOS_PASSWORD": "<IVORYOS_PASSWORD>"
      }
    }
  }
}

📦 Installation

Install uv.

1. Clone the Repository

git clone https://gitlab.com/heingroup/ivoryos-mpc
cd ivoryos-mcp

2. Install dependencies

When using IDE (e.g. PyCharm), the uv environment might be configured, you can skip this section.

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install -r uv.lock

⚙️ Configuration

Option 1: in .env, change ivoryOS url and login credentials.

IVORYOS_URL=http://127.0.0.1:8000/ivoryos
IVORYOS_USERNAME=admin
IVORYOS_PASSWORD=admin

Option 2: In ivoryos_mcp/server.py, change ivoryOS url and login credentials.

url = "http://127.0.0.1:8000/ivoryos"
login_data = {
    "username": "admin",
    "password": "admin",
}

🚀 Install the server (in Claude Desktop)

mcp install ivoryos_mcp/server.py

✨ Features

Category	Feature	Route	Description
ℹ️ General Tools	platform-info	GET /instruments	Get ivoryOS info and signature of the platform
	execution-status	GET /executions/status	Check if system is busy and current/last task status
ℹ️ Workflow Design	list-workflow-scripts	GET /library/<deck_name>	List all workflow scripts from the database
	load-workflow-script	GET /library/<workflow_name>	Load a workflow script from the database
	submit-workflow-script	POST /draft/submit_python	Save a workflow Python script to the database
	get-optmizer-schema	POST /executions/optimizer_schema	Get all optimizers schema
ℹ️ Workflow Data	list-workflow-data	GET /executions/records	List available workflow execution data
	load-workflow-data	GET /executions/records/<workflow_id>	Load execution log and data file
🤖 Direct Control	execute-task	POST /instruments/<component>	Call platform function directly
🤖 Workflow Run	run-workflow-repeat	POST /executions/config	Run workflow scripts repeatedly with static parameters
	run-workflow-kwargs	POST /executions/config	Run workflow scripts with dynamic parameters
	run-workflow-campaign	POST /executions/campaign	Run workflow campaign with an optimizer
🤖 Workflow Control	pause-and-resume	GET /executions/pause-resume	Pause or resume the workflow execution
	abort-pending-workflow	GET /executions/abort/next-iteration	Finish current iteration, abort future executions
	stop-current-workflow	GET /executions/abort/next-task	Safe stop of current workflow

⚠️ ℹ️ are resources, but decorated as tool due to the current issue with MCP Python SDK and Claude Desktop integration.
It's recommended to only use allow always for ℹ️ tasks
and use allow once for 🤖 tasks. These tasks will trigger actual actions on your hosted Python code.

🧪 Examples

The example prompt uses the abstract SDL example.

Platform info

Load prebuilt workflow script