Arize Tracing Assistant

Warning

Deprecated: This MCP server is deprecated and no longer maintained by Arize. It will receive no new features, fixes, or security updates. Use the Arize AX MCP Server instead to give your coding agent access to your Arize AX projects, traces, datasets, experiments, evaluators, monitors, and prompts, and Arize Skills for supported, end-to-end Arize AX workflows.

Overview

This legacy MCP server provides documentation and examples for instrumenting AI applications with Arize AX. It has been superseded by the Arize AX MCP Server, a hosted remote server that connects with an endpoint URL and an Arize API key and needs no local installation. For maintained Arize AX workflows, install and use Arize Skills.

Legacy installation

Install uv

Make sure uv - the fast Python package manager - is installed on your system. Installation instructions: uv installation guide.

• On macOS:

  pip install uv

  or

  brew install uv

• On Linux:

  curl -LsSf https://astral.sh/uv/install.sh | less

• On Windows:

  powershell -c "irm https://astral.sh/uv/install.ps1 | more"

IDE integration

Antigravity

1. Open the MCP Store panel within the "..." dropdown at the top of the editor's side panel.
2. Type "Arize" in the search and click Install.

OR

1. To Install as a Custom MCP Server, Click on "Manage MCP Servers"
2. Click on "View raw config"
3. Add the following code to your mcp_config.json:

  "mcpServers": {
    "arize-tracing-assistant": {
      "command": "uvx",
      "args": [
      "arize-tracing-assistant@latest"
    ]
    }
  }

Cursor

1. Go to Cursor Settings -> MCP.
2. Click "Add new global MCP server" and add the server to your config JSON.
3. Remove the env section if you don't have access to RunLLM.

Example config:

"arize-tracing-assistant": {
  "command": "uvx",
  "args": [
    "arize-tracing-assistant@latest"
  ]
}

---

Claude Desktop

1. Go to Claude Desktop Settings.
2. In Developer > Edit Config and add this to your config JSON:

"arize-tracing-assistant": {
  "command": "/Users/myuser/miniconda3/bin/uvx",
  "args": [
   "arize-tracing-assistant@latest"
  ]
}

---

Manual MCP config

Add the following snippet to your MCP config file:

  "mcpServers": {
    "arize-tracing-assistant": {
      "command": "uvx",
      "args": [
      "arize-tracing-assistant@latest"
    ]
    }
  }

CLI Integration

Gemini CLI

Install using the Gemini CLI extension:

gemini extensions install https://github.com/Arize-ai/arize-tracing-assistant

Usage

Once the MCP server is running, ask your IDE or LLM questions about tracing or general Arize support like...

• Instrument this app using Arize
• Can you use manual instrumentation so that I have more control over my traces?
• How can I redact sensitive information from my spans?
• Use decorators to have more control over the span attributes
• Can you make sure the context of this trace is propagated across these tool calls?
• Where can I find my Arize keys?

Troubleshooting

• Make sure the JSON configs are perfectly formatted.

• Clear the uv cache with uv cache clean to force accessing the latest version.

• Make sure your uv command is pointing to the right location by running which uv, or simply use the full path.

• The server should start in the terminal just by running:

  uvx arize-tracing-assistant@latest

• Use the Anthropic MCP inspector by running:

  npx @modelcontextprotocol/inspector uvx arize-tracing-assistant@latest