Arize Phoenix

Arize Phoenix

MCP server implementation for Arize Phoenix providing unified interface to Phoenix's capabilities

7,778
Stars
637
Forks
20
Releases

Overview

The Arize Phoenix MCP (Model Context Protocol) server provides a unified API that exposes Phoenix's capabilities through MCP, enabling standardized access to tracing, evaluation, datasets, experiments, and prompt management. It is vendor- and language-agnostic as Phoenix itself is, and is delivered as a TypeScript server package (phoenix-mcp) that can be used with the broader Phoenix ecosystem, including Python and TypeScript subpackages. The MCP server can be deployed in various environments (local, notebooks, containerized deployments, or cloud) and interoperates with OpenTelemetry-based tracing and other Phoenix integrations. It is designed for developers and teams building MCP-based integrations who want a single API to coordinate model context, prompts, and observability workflows across Phoenix components. By providing a unified interface to Phoenix's capabilities, the MCP server enables versioned prompts, coordinated experiments, and consistent interaction patterns for tracing, evaluation, and prompt-management workflows.

Details

Owner
Arize-ai
Language
Jupyter Notebook
License
Other
Updated
2025-12-07

Audience

DevelopersProvide MCP-based access to Arize Phoenix capabilities via a unified API.

Tags

MCPModel Context ProtocolPhoenixobservabilitytracingdatasetsexperimentspromptsprompt management