playbooks

Home / MCP / Phoenix MCP Server

Phoenix MCP Server

Provides programmatic access to Phoenix projects, spans, prompts, datasets, and experiments via MCP.

typescript
7.9kstars
View on GitHub
Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
    "mcpServers": {
        "phoenix": {
            "command": "npx",
            "args": [
                "-y",
                "@arizeai/phoenix-mcp@latest",
                "--baseUrl",
                "https://my-phoenix.com",
                "--apiKey",
                "your-api-key"
            ],
            "env": {
                "PHOENIX_API_KEY": "YOUR_PHOENIX_API_KEY",
                "PHOENIX_BASE_URL": "https://my-phoenix.com"
            }
        }
    }
}

Arize Phoenix MCP Server implements the Model Context Protocol for the Arize Phoenix platform, providing a unified interface to Phoenix capabilities for managing projects, spans and annotations, prompts, datasets, and experiments. You can connect your client to this MCP server to retrieve data, manage resources, and run iterative prompts against Phoenix through a standardized interface.

How to use

You interact with the Phoenix MCP Server through an MCP client by configuring a connection that points to the server you want to query. Use the client to list projects, fetch spans and their annotations, manage prompts, explore datasets, and pull experiment results. The server exposes capabilities that let you discover data sources, retrieve associated metadata, and iterate on prompts with the help of an LLM.

To start, configure your MCP client to connect to the Phoenix MCP Server using the provided runtime command. You will typically supply your Phoenix base URL and API key, and the client will send requests to Phoenix to retrieve data, prompts, datasets, and experiments.

How to install

Prerequisites: Node.js (for running the MCP via npx) and a package manager. The setup example uses pnpm for development workflows.

// From the project root where the MCP is consumed
// Install dependencies and build if necessary. The following commands match the development approach described:

pnpm install
pnpm build

Additional configuration and usage notes

The Phoenix MCP Server can be run using a runtime command that leverages npx to fetch the latest Phoenix MCP package, supplying your base URL and API key as arguments.

{
  "mcpServers": {
    "phoenix": {
      "command": "npx",
      "args": [
        "-y",
        "@arizeai/phoenix-mcp@latest",
        "--baseUrl",
        "https://my-phoenix.com",
        "--apiKey",
        "your-api-key"
      ]
    }
  }
}

Environment variables

When developing or running in a local environment, you need to provide Phoenix API credentials. Set these values in a .env file or your environment: - PHOENIX_API_KEY: Your Phoenix API key - PHOENIX_BASE_URL: The base URL for Phoenix

Development and community

Developers can contribute to Phoenix MCP Server and related client libraries. You can join the community for support, read model context guides, and participate in discussions about roadmap and features.

License

Apache 2.0

Available tools

projects

List and explore projects that organize your observability data

spans

Retrieve spans and their annotations for analysis and debugging

prompts

Create, list, update, and iterate on prompts

datasets

Explore datasets and synthesize new examples

experiments

Pull experiment results and visualize them with an LLM