Developer Tools stdio

Phoenix

Name: Phoenix
Author: Arize-ai

Created by Arize-ai

Provides an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting of LLM applications.

How to Connect & Install

Select your client or environment below to copy the verified, ready-to-run installation configuration.

1-line autonomous agent setup instruction. Paste into Claude Code, Codex CLI, or Cursor Agent:

Fetch and configure Model Context Protocol (MCP) server Phoenix from https://www.fastmcp.dev/server/phoenix

Agents automatically discover, install, and execute the server spec without manual token setup.

Connect via direct Stateless MCP v2 / Streamable HTTP endpoint or JSON-RPC:

https://github.com/arize-ai/phoenix

Stateless edge endpoint compliant with MCP 2024-11-05 and v2 draft protocol specs.

Includes server instructions and tool definitions for Claude:

claude mcp add phoenix -- npx -y phoenix

Run in terminal to register server in current project or pass --scope user for global availability.

Configures server instructions and tools for OpenAI Codex & ChatGPT Desktop:

[mcp_servers.phoenix]
command = "npx"
args = ["-y","phoenix"]

Applies to ChatGPT Desktop app, Codex CLI, and IDE extension across your Codex host.

Add to Cursor project settings (.cursor/mcp.json) or Cursor Settings > Features > MCP:

{
  "mcpServers": {
    "phoenix": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "phoenix"
      ]
    }
  }
}

Commit to version control to allow teammates to connect immediately.

Install as an autonomous Agent Skill using the standard skills CLI:

npx skills add https://github.com/arize-ai/phoenix

Standard agent skill installer compatible with Claude Code, Cursor, and agy harnesses.

Server Guidance & Instructions OpenAI & Claude Spec Compliant

Phoenix: Provides an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting of LLM applications. Key capabilities: Leverage LLMs to benchmark application performance using response and retrieval evals.; Track and evaluate changes to prompts, LLMs, and retrieval.

Codex and Claude read this instruction prompt during initialization to orchestrate cross-tool workflows and rate limits.

Top Core Functions

01 Trace LLM application runtime execution

02 Benchmark application performance using evals

03 Manage datasets for fine tuning