Connecting AI Agents to JEDI with Built-In MCP Server

JEDI includes a built-in MCP server that enables AI agents to connect to your JEDI instance for real-time data access, historical analysis, and dashboard creation.

Key Capabilities of JEDI’s Built-In MCP Server

Direct AI Agent Connectivity

Includes SSE MCP server for Claude Code, Gemini CLI, and OpenAI Codex, and MCP Gateway bridge for STDIO-based clients like Claude Desktop.
Real-Time Metric Discovery

AI agents can discover and search available device metrics within your JEDI environment.
Real-Time Monitoring

AI agents can retrieve the latest values from connected sensors and devices in real time.
Historical Data Analysis

The MCP server enables AI agents to query historical sensor data.
AI-Assisted Dashboard Creation

Allows AI agents to create dashboards directly inside the JEDI webapps environment.
Secure Scoped File Operations

Provides a controlled environment for AI-generated dashboard workflows with secure dashboard file operations.

Configuring Your AI Agent

To allow an AI agent to communicate with JEDI, you must configure the agent to connect to the JEDI MCP endpoint.

The default MCP endpoint is:

http://localhost:9124/mcp

Step 1: Start JEDI

Launch the JEDI application. When successful, JEDI automatically starts the MCP server on port 9124.

Step 2: Configure Your AI Agent

Choose the appropriate configuration based on your AI agent.

Claude Code: Create .mcp.json in your project root:

{
  "mcpServers": {
    "jedi": {
      "type": "sse",
      "url": "http://localhost:9124/mcp"
    }
  }
}

Gemini CLI: Create .gemini/settings.json in your project root:

{
  "mcpServers": {
    "jedi": {
      "type": "sse",
      "url": "http://localhost:9124/mcp"
    }
  }
}

OpenAI Codex: Configure ~/.codex/config.toml with:

[mcp_servers.jedi]
url = "http://localhost:9124/mcp"

Claude Desktop: Claude Desktop requires the included MCP Gateway because it does not support SSE transport directly.

Configure claude_desktop_config.json with:

{
  "mcpServers": {
    "jedi": {
      "command": "<path-to-JEDI>\\tools\\mcp\\mcp-gateway.exe",
      "args": ["-server", "http://localhost:9124/mcp"]
    }
  }
}

Note: Replace <path-to-JEDI> with the actual path to your JEDI folder.

Step 3: Restart Your AI Agent

Restart or reload your AI agent after updating the configuration.

Step 4: Verify the Connection

Ask the AI agent:

List all available metrics

If configured correctly, the AI agent will return metrics discovered from your JEDI instance.

Example: Creating a Dashboard

Once connected, you can use natural language prompts to instruct your AI agent to create dashboards using JEDI’s built-in dashboard tools.

Example prompts:

Create a temperature monitoring dashboard
Build a real-time environmental monitoring dashboard
Create a dashboard showing humidity and pressure trends for the last 24 hours

Note

Refresh JEDI on browser to see the new dashboard.

Troubleshooting

Cannot Connect to JEDI

Verify that JEDI is running and the MCP server is active.

Test the endpoint:

curl http://localhost:9124/mcp

A valid SSE response confirms the MCP server is available.

Port Already in Use

If port 9124 is occupied:

Change the MCP port in jedi.conf

[AppConfig]
    mcp_port = "7101"

Update your AI agent configuration
Restart JEDI and the AI agent

No Metrics Returned

If the AI agent cannot find metrics:

Open the JEDI UI at http://localhost:9123
Verify data collectors are enabled
Confirm sensors are sending data

Dashboard Creation Issues

If generated dashboards do not render correctly:

Confirm the HTTP data collector is enabled
Verify the webapps directory is writable
Restart the AI agent after restarting JEDI
Review JEDI logs for permission or file write errors

Begin building intelligent, AI-assisted monitoring experiences with JEDI!

Now that your AI agent is connected, continue exploring dashboard creation workflows with real-time and historical data access.

Need help?

Contact our support team: support@machinechat.io