Skip to main content
Glama
Opik MCP Server

Opik MCP Server

by comet-ml

执照 Node.js 版本 TypeScript

⚠️ 注意: SSE(服务器发送事件)传输支持目前处于实验阶段,尚未经过测试。对于生产环境,我们建议使用 IDE 集成示例中所示的直接流程执行方法。

🚀 什么是 Opik MCP 服务器?

Opik MCP 服务器是 Opik 平台模型上下文协议 (MCP) 的开源实现。它提供了与 Opik 功能交互的统一接口,支持多种传输机制,可灵活集成到各种环境中。

您可以使用 Opik MCP 服务器进行以下操作:

  • IDE 集成:

    • 与 Cursor 和其他兼容 IDE 无缝集成

    • 可以从您的开发环境直接访问 Opik 的功能

  • 统一 API 访问:

    • 通过标准化协议访问所有 Opik 功能

    • 针对不同的集成场景利用多种传输选项(stdio、SSE)

  • 平台管理:

    • 通过一致的界面管理提示、项目、跟踪和指标

    • 高效组织和监控你的法学硕士 (LLM) 申请

Related MCP server: dap-mcp

特征

  • 提示管理:创建、列出、更新和删除提示

  • 项目/工作区管理:组织和管理项目

  • 跟踪:跟踪和分析跟踪数据

  • 指标:收集和查询指标数据

快速入门

安装

手动安装

# Clone the repository
git clone https://github.com/comet-ml/opik-mcp.git
cd opik-mcp

# Install dependencies and build
npm install
npm run build

配置

根据示例创建一个.env文件:

cp .env.example .env
# Edit .env with your specific configuration

启动服务器

# Start with stdio transport (default)
npm run start:stdio

# Start with SSE transport for network access (experimental)
npm run start:sse

IDE 集成

光标集成

要与 Cursor IDE 集成,请在项目目录中创建一个.cursor/mcp.json文件,并采用以下配置:

{
  "mcpServers": {
    "opik": {
      "command": "/path/to/node",
      "args": [
        "/path/to/opik-mcp/build/index.js",
        "--apiUrl",
        "https://www.comet.com/opik/api",
        "--apiKey",
        "YOUR_API_KEY",
        "--workspace",
        "default",
        "--debug",
        "true"
      ],
      "env": {
        "OPIK_API_BASE_URL": "https://www.comet.com/opik/api",
        "OPIK_API_KEY": "YOUR_API_KEY",
        "OPIK_WORKSPACE_NAME": "default",
      }
    }
  }
}

将/path/to/node替换为 Node.js 可执行文件的路径,将/path/to/opik-mcp替换为 opik-mcp 安装的路径。同时将YOUR_API_KEY替换为你实际的 Opik API 密钥。

可用命令

该项目包含用于常见操作的 Makefile:

# Display all available commands
make help

# Run tests
make test

# Run transport-specific tests
make test-transport

# Start the server with SSE transport (experimental)
make start-sse

# Start the server with stdio transport
make start-stdio

交通选择

标准输入/输出

非常适合客户端和服务器在同一台机器上运行的本地集成。

make start-stdio

服务器发送事件 (SSE)

通过 HTTP 启用远程访问和多个客户端同时访问。请注意,此传输选项尚处于实验阶段。

make start-sse

有关 SSE 传输的详细信息,请参阅docs/sse-transport.md 。

发展

测试

# Run all tests
npm test

# Run specific test suite
npm test -- tests/transports/sse-transport.test.ts

预提交钩子

该项目使用预提交钩子来确保代码质量:

# Run pre-commit checks manually
make precommit

文档

执照

Apache 2.0

Available Tools

19 tools
create-projectCInspect

Create a new project/workspace

Inputs: name (required), description, workspaceName.

create-promptCInspect

Create a new prompt

Inputs: name (required).

create-prompt-versionCInspect

Create a new version of a prompt

Inputs: commit_message (required), name (required), template (required).

delete-projectCInspect

Delete a project

Inputs: projectId (required), workspaceName.

delete-promptCInspect

Delete a prompt

Inputs: promptId (required).

get-metricsCInspect

Get metrics data

Inputs: endDate, metricName, projectId, projectName, +1 more.

get-opik-examplesCInspect

Get examples of how to use Opik Comet's API for specific tasks

Inputs: task (required).

get-opik-helpBInspect

Get contextual help about Opik Comet's capabilities

Inputs: topic (required), subtopic.

get-opik-tracing-infoCInspect

Get information about Opik's tracing capabilities and how to use them

Inputs: topic.

get-project-by-idCInspect

Get a single project by ID

Inputs: projectId (required), workspaceName.

get-prompt-by-idCInspect

Get a single prompt by ID

Inputs: promptId (required).

get-server-infoBInspect

Get information about the Opik server configuration

Inputs: random_string.

get-trace-by-idCInspect

Get a single trace by ID

Inputs: traceId (required), workspaceName.

get-trace-statsCInspect

Get statistics for traces

Inputs: endDate, projectId, projectName, startDate, +1 more.

list-projectsCInspect

Get a list of projects/workspaces

Inputs: page (required), size (required), sortBy, sortOrder, +1 more.

list-promptsCInspect

Get a list of Opik prompts

Inputs: page (required), size (required).

list-tracesCInspect

Get a list of traces

Inputs: page (required), size (required), projectId, projectName, +1 more.

update-projectCInspect

Update a project

Inputs: projectId (required), description, name, workspaceName.

update-promptCInspect

Update a prompt

Inputs: name (required), promptId (required).

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    B3.1/5.0

    Scored across 19 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose targeting specific resources (projects, prompts, traces, metrics, help) and actions (create, get, list, update, delete). There is no overlap or ambiguity between tools, making it easy for an agent to select the right one for any operation.

    Naming Consistency5/5

    All tools follow a consistent verb_noun pattern with hyphens (e.g., create-project, get-project-by-id, list-projects). The naming is uniform across all 19 tools, using clear verbs like create, get, list, update, and delete paired with specific nouns.

    Tool Count4/5

    With 19 tools, the count is slightly high but reasonable for a comprehensive server covering projects, prompts, traces, metrics, and help. It feels well-scoped for the Opik domain, though it borders on being heavy compared to typical 3-15 tool ranges.

    Completeness5/5

    The tool set provides complete CRUD/lifecycle coverage for projects and prompts (create, get, list, update, delete), with additional tools for traces, metrics, and help. There are no obvious gaps; agents can perform all core operations without dead ends in this domain.

    Maintenance

    ActivityActive
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      B
      quality
      D
      maintenance
      Implements the Model Context Protocol (MCP) to provide AI models with a standardized interface for connecting to external data sources and tools like file systems, databases, or APIs.
      1
      153
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      An implementation of the Model Context Protocol (MCP) that enables interaction with debug adapters, allowing language models to control debuggers, set breakpoints, evaluate expressions, and navigate source code during debugging sessions.
      41
      AGPL 3.0
    • F
      license
      Not graded
      quality
      D
      maintenance
      A Python-based implementation of the Model Context Protocol that enables communication between a model context management server and client through a request-response architecture.
      -
    • -
      license
      Not graded
      quality
      D
      maintenance
      A standardized foundation for building Model Context Protocol servers that integrate with VS Code, using Python with stdio transport for seamless AI tool integration.
      -