Skip to main content
Glama
Sign In
deenesjakoruzh
@arizeai/phoenix-mcp

@arizeai/phoenix-mcp

Official
by Arize-ai
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

Phoenix MCP Server is an implementation of the Model Context Protocol for the Arize Phoenix platform. It provides a unified interface to Phoenix's capabilities.

You can use Phoenix MCP Server for:

  • Projects Management: List and explore projects that organize your observability data

  • Traces, Spans & Annotations: Retrieve traces, spans, and annotation configs for analysis and debugging

  • Sessions: Explore conversation flows and session-level annotations

  • Prompts Management: Create, list, update, and iterate on prompts

  • Datasets: Explore datasets and synthesize new examples

  • Experiments: Pull experiment results and visualize them with the help of an LLM

Don't see a use-case covered? @arizeai/phoenix-mcp is open-source! Issues and PRs welcome.

Installation

This MCP server can be used with npx and can be directly integrated with clients like Claude Desktop, Cursor, and more.

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

Related MCP server: @prompthub/mcp

Development

Install

This package is managed via a pnpm workspace.

// From the /js/ directory
pnpm install
pnpm build

This only needs to be repeated if dependencies change or there is a change to the phoenix-client.

Building

To build the project:

pnpm build

Development Mode

To run in development mode:

pnpm dev

Debugging

You can build and run the MCP inspector using the following:

pnpm inspect

Environment Variables

When developing, the server requires the following environment variables:

  • PHOENIX_API_KEY: Your Phoenix API key

  • PHOENIX_ENDPOINT: The base URL for Phoenix

  • PHOENIX_PROJECT: Optional default project for project-scoped tools (alias: PHOENIX_PROJECT_NAME)

  • PHOENIX_CLIENT_HEADERS: Optional JSON-encoded request headers

Make sure to set these in a .env file. See .env.example.

At runtime the server also discovers the nearest .env.phoenix file at or above the working directory. Process environment variables and command-line options take precedence. If a higher-priority API key is paired with an endpoint from the file, the server warns once and continues. Set PHOENIX_DISCOVER_CONFIG=false to disable discovery.

Tool Coverage

The MCP server covers the main operational Phoenix workflows:

Prompts — list-prompts, get-prompt, get-latest-prompt, get-prompt-by-identifier, get-prompt-version, list-prompt-versions, get-prompt-version-by-tag, list-prompt-version-tags, add-prompt-version-tag, upsert-prompt

Projects — list-projects, get-project

Traces — list-traces, get-trace

Spans — get-spans, get-span-annotations

Sessions — list-sessions, get-session

Annotation Configs — list-annotation-configs

Datasets — list-datasets, get-dataset, get-dataset-examples, get-dataset-experiments, add-dataset-examples

Experiments — list-experiments-for-dataset, get-experiment-by-id

For Phoenix documentation search, use the separate Phoenix Docs MCP server instead of this package.

Community

Join our community to connect with thousands of AI builders:

License

Apache 2.0

Available Tools

27 tools
add-dataset-examplesA

Add examples to an existing dataset.

This tool adds one or more examples to an existing dataset. Each example includes an input, output, and metadata. The metadata will automatically include information indicating that these examples were synthetically generated via MCP. When calling this tool, check existing examples using the "get-dataset-examples" tool to ensure that you are not adding duplicate examples and following existing patterns for how data should be structured.

Example usage: Look at the analyze "my-dataset" and augment them with new examples to cover relevant edge cases

Expected return: Confirmation of successful addition of examples to the dataset.

Inputs: examples (required), dataset_name (required).

add-prompt-version-tagA

Add a tag to a specific prompt version. The operation returns no content on success (204 status code).

Example usage: Tag prompt version 'promptversionid1234' with the name 'production'

Expected return: Confirmation message of successful tag addition

Inputs: name (required), prompt_version_id (required), description.

get-datasetC

Get dataset metadata by name or ID.

Example usage: Show me the dataset "my-dataset"

Expected return: A dataset object with metadata and version information.

Inputs: dataset_id, dataset_name.

get-dataset-examplesB

Get examples from a dataset.

Dataset examples are an array of objects that each include an input, (expected) output, and optional metadata. These examples are typically used to represent input to an application or model (e.g. prompt template variables, a code file, or image) and used to test or benchmark changes.

Example usage: Show me all examples from dataset RGF0YXNldDox

Expected return: Object containing dataset ID, version ID, and array of examples.

Inputs: splits, dataset_id, version_id, dataset_name.

get-dataset-experimentsC

List experiments run on a dataset.

Example usage: Show me all experiments run on dataset RGF0YXNldDox

Expected return: Array of experiment objects with metadata.

Inputs: limit, dataset_id, dataset_name.

get-experiment-by-idA

Get an experiment by its ID.

The tool returns experiment metadata in the first content block and a JSON object with the experiment data in the second. The experiment data contains both the results of each experiment run and the annotations made by an evaluator to score or label the results, for example, comparing the output of an experiment run to the expected output from the dataset example.

Example usage: Show me the experiment results for experiment RXhwZXJpbWVudDo4

Expected return: Object containing experiment metadata and results.

Inputs: experiment_id (required).

get-latest-promptA

Get the latest version of a prompt. Returns the prompt version with its template, model configuration, and invocation parameters.

Example usage: Get the latest version of a prompt named 'article-summarizer'

Expected return: Prompt version object with template and configuration. Example: { "description": "Initial version", "model_provider": "OPENAI", "model_name": "gpt-3.5-turbo", "template": { "type": "chat", "messages": [ { "role": "system", "content": "You are an expert summarizer. Create clear, concise bullet points highlighting the key information." }, { "role": "user", "content": "Please summarize the following {{topic}} article:\n\n{{article}}" } ] }, "template_type": "CHAT", "template_format": "MUSTACHE", "invocation_parameters": { "type": "openai", "openai": {} }, "id": "promptversionid1234" }

Inputs: prompt_identifier (required).

get-projectC

Get a project by name or ID.

Example usage: Show me the project "default"

Expected return: A single project object with metadata.

Inputs: project_identifier (required).

get-promptC

Get a prompt using a single MCP-native interface.

Provide a prompt identifier to fetch the latest version, or add a tag or versionId to select a specific version.

Example usage: Get prompt "article-summarizer" Get prompt "article-summarizer" with tag "production" Get prompt "article-summarizer" using version "promptversionid1234"

Expected return: Prompt version object with template and configuration.

Inputs: prompt_identifier (required), tag, version_id.

get-prompt-by-identifierA

Get a prompt's latest version by its identifier (name or ID). Returns the prompt version with its template, model configuration, and invocation parameters.

Example usage: Get the latest version of a prompt with name 'article-summarizer'

Expected return: Prompt version object with template and configuration.

Inputs: prompt_identifier (required).

get-prompt-versionA

Get a specific version of a prompt using its version ID. Returns the prompt version with its template, model configuration, and invocation parameters.

Example usage: Get a specific prompt version with ID 'promptversionid1234'

Expected return: Prompt version object with template and configuration.

Inputs: prompt_version_id (required).

get-prompt-version-by-tagA

Get a prompt version by its tag name. Returns the prompt version with its template, model configuration, and invocation parameters.

Example usage: Get the 'production' tagged version of prompt 'article-summarizer'

Expected return: Prompt version object with template and configuration.

Inputs: tag_name (required), prompt_identifier (required).

get-sessionA

Get a single session by GlobalID or user-provided session_id.

Example usage: Show me session "chat-123"

Expected return: A session object and, optionally, its annotations.

Inputs: session_identifier (required), include_annotations.

get-span-annotationsA

Get span annotations for a list of span IDs.

Span annotations provide additional metadata, scores, or labels for spans. They can be created by humans, LLMs, or code and help in analyzing and categorizing spans.

Example usage: Get annotations for spans ["span1", "span2"] from project "my-project" Get quality score annotations for span "span1" from project "my-project"

Expected return: Object containing annotations array and optional next cursor for pagination. Example: { "annotations": [ { "id": "annotation123", "span_id": "span1", "name": "quality_score", "result": { "label": "good", "score": 0.95, "explanation": null }, "annotator_kind": "LLM", "metadata": { "model": "gpt-4" } } ], "nextCursor": "cursor_for_pagination" }

Inputs: span_ids (required), limit, cursor, project_identifier, +2 more.

get-spansB

Get spans from a project with filtering criteria.

Spans represent individual operations or units of work within a trace. They contain timing information, attributes, and context about the operation being performed.

Example usage: Get recent spans from project "my-project" Get spans in a time range from project "my-project"

Expected return: Object containing spans array and optional next cursor for pagination. Example: { "spans": [ { "id": "span123", "name": "http_request", "context": { "trace_id": "trace456", "span_id": "span123" }, "start_time": "2024-01-01T12:00:00Z", "end_time": "2024-01-01T12:00:01Z", "attributes": { "http.method": "GET", "http.url": "/api/users" } } ], "nextCursor": "cursor_for_pagination" }

Inputs: limit, names, cursor, end_time, +7 more.

get-traceA

Get a single trace by its exact trace ID within a project.

Example usage: Show me trace "abc123def456" from project "default"

Expected return: A trace object with all spans that belong to the trace.

Inputs: trace_id (required), project_identifier, include_annotations.

list-annotation-configsB

List Phoenix annotation configs.

Annotation configs define the available human or automated labels, scores, and freeform annotation types.

Example usage: Show me all annotation configs

Expected return: Array of annotation config objects.

Inputs: limit.

list-datasetsA

Get a list of all datasets.

Datasets are collections of 'dataset examples' that each example includes an input, (expected) output, and optional metadata. They are primarily used as inputs for experiments.

Example usage: Show me all available datasets

Expected return: Array of dataset objects with metadata. Example: [ { "id": "RGF0YXNldDox", "name": "my-dataset", "description": "A dataset for testing", "metadata": {}, "created_at": "2024-03-20T12:00:00Z", "updated_at": "2024-03-20T12:00:00Z" } ]

Inputs: limit.

list-experiments-for-datasetB

Get a list of all the experiments run on a given dataset.

Experiments are collections of experiment runs, each experiment run corresponds to a single dataset example. The dataset example is passed to an implied task which in turn produces an output.

Example usage: Show me all the experiments I've run on dataset RGF0YXNldDox

Expected return: Array of experiment objects with metadata. Example: [ { "id": "experimentid1234", "dataset_id": "datasetid1234", "dataset_version_id": "datasetversionid1234", "repetitions": 1, "metadata": {}, "project_name": "Experiment-abc123", "created_at": "YYYY-MM-DDTHH:mm:ssZ", "updated_at": "YYYY-MM-DDTHH:mm:ssZ" } ]

Inputs: limit, dataset_id, dataset_name.

list-projectsC

Get a list of all projects.

Projects are containers for organizing traces, spans, and other observability data. Each project has a unique name and can contain traces from different applications or experiments.

Example usage: Show me all available projects

Expected return: Array of project objects with metadata. Example: [ { "id": "UHJvamVjdDox", "name": "default", "description": "Default project for traces" }, { "id": "UHJvamVjdDoy", "name": "my-experiment", "description": "Project for my ML experiment" } ]

Inputs: limit, cursor, include_experiment_projects.

list-promptsC

Get a list of all the prompts.

Prompts (templates, prompt templates) are versioned templates for input messages to an LLM. Each prompt includes both the input messages, but also the model and invocation parameters to use when generating outputs.

Returns a list of prompt objects with their IDs, names, and descriptions.

Example usage: List all available prompts

Expected return: Array of prompt objects with metadata. Example: [{ "name": "article-summarizer", "description": "Summarizes an article into concise bullet points", "source_prompt_id": null, "id": "promptid1234" }]

Inputs: limit.

list-prompt-versionsB

Get a list of all versions for a specific prompt. Returns versions with pagination support.

Example usage: List all versions of a prompt named 'article-summarizer'

Expected return: Array of prompt version objects with IDs and configuration.

Inputs: prompt_identifier (required), limit.

list-prompt-version-tagsB

Get a list of all tags for a specific prompt version. Returns tag objects with pagination support.

Example usage: List all tags associated with prompt version 'promptversionid1234'

Expected return: Array of tag objects with names and IDs.

Inputs: prompt_version_id (required), limit.

list-sessionsB

List sessions for a project.

Sessions represent conversation flows grouped across traces.

Example usage: Show me the last 10 sessions for project "default"

Expected return: Array of session objects ordered by the requested sort order.

Inputs: limit, order, project_identifier.

list-tracesB

List traces for a project.

This tool groups project spans into traces and returns the newest traces first.

Example usage: Show me the last 10 traces for project "default" Show me recent traces from the last 30 minutes for project "checkout"

Expected return: Array of trace objects with grouped spans and summary timing information.

Inputs: limit, since, last_n_minutes, project_identifier, +1 more.

phoenix-supportA

Get help with Phoenix and OpenInference.

  • Tracing AI applications via OpenInference and OpenTelemetry

  • Phoenix datasets, experiments, and prompt management

  • Phoenix evals and annotations

Use this tool when you need assistance with Phoenix features, troubleshooting, or best practices.

Expected return: Expert guidance about how to use and integrate Phoenix

Inputs: query (required).

upsert-promptC

Create or update a prompt with its template and configuration. Creates a new prompt and its initial version with specified model settings.

Example usage: Create a new prompt named 'email_generator' with a template for generating emails

Expected return: A confirmation message of successful prompt creation

Inputs: name (required), template (required), model_name, description, +2 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

    Deploy Server
    One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

    TDQS

    B3.2/5.0

    Scored across 27 tools

    Disambiguation3/5

    Multiple tools for fetching prompts (get-prompt-by-identifier, get-prompt, get-latest-prompt) have overlapping functionality, which could cause confusion. Other tools are distinct.

    Naming Consistency4/5

    Tool names follow a consistent verb-noun pattern with hyphens (e.g., list-prompts, get-dataset). Minor deviations like get-latest-prompt vs get-prompt are present but overall consistent.

    Tool Count4/5

    27 tools cover multiple domains (prompts, datasets, experiments, traces, etc.) and are justified given the breadth of functionality, though slightly on the high side.

    Completeness3/5

    The tool surface covers core operations for prompts, datasets, and observability, but lacks delete operations for prompts and datasets, and does not include create dataset or project.

    Maintenance

    ActivityActive
    ResponsivenessSyncing

    Resources

    Related MCP Connectors

    Related MCP Servers

    • MCP Server Boilerplate

      cds-id
      A
      license
      D
      quality
      D
      maintenance
      A ready-to-use starter implementation of the Model Context Protocol (MCP) server that enables applications to provide standardized context for LLMs with sample resources, tools, and prompts.
      2
      5 npm
      1
      MIT
    • synapse-mcp

      synapse-mcpofficial

      Sage-Bionetworks
      A
      license
      Not graded
      quality
      B
      maintenance
      A Model Context Protocol (MCP) server that enables AI agent access to Synapse entities such as Datasets, Projects, Folders, Files, Tables, and more.
      4
      MIT
    • picsellia-mcp

      picsellia-mcpofficial

      picselliahq
      F
      license
      Not graded
      quality
      D
      maintenance
      MCP server that wraps the Picsellia Python SDK, giving AI assistants access to computer vision platform capabilities such as datasets, experiments, models, deployments, and monitoring.
      -