phoenix

Overview

MCP server for Arize Phoenix observability platform

Transport

stdio

Tools

• add-dataset-examples
• add-prompt-version-tag
• get-dataset-examples
• get-dataset-experiments
• get-experiment-by-id
• get-latest-prompt
• get-prompt-by-identifier
• get-prompt-version
• get-prompt-version-by-tag
• get-span-annotations
• get-spans
• list-datasets
• list-experiments-for-dataset
• list-projects
• list-prompt-version-tags
• list-prompt-versions
• list-prompts
• phoenix-support
• upsert-prompt

Key Capabilities

• Dataset and experiment management for LLM evaluation workflows
• Prompt version control and tagging for LLM application development
• Span and trace querying for observability and debugging
• Project-based organization for managing multiple Phoenix instances

How It Works

This MCP server provides AI assistants with programmatic access to Arize Phoenix, an open-source observability platform for LLMs. It enables:

1. Creating and managing evaluation datasets with examples
2. Tracking and versioning prompts with tags for reproducibility
3. Querying spans, traces, and annotations for LLM application monitoring
4. Running experiments and tracking their results across different datasets