opik
Model Context Protocol (MCP) implementation for Opik enabling seamless IDE integration and unified access to prompts, projects, traces, and metrics.
claude mcp add --transport stdio comet-ml-opik-mcp npx -y opik-mcp --apiKey YOUR_API_KEY
How to use
The Opik MCP Server exposes a standard Model Context Protocol (MCP) interface for interacting with Comet Opik workflows. It supports both local stdio transport for on-device integrations and a remote streamable-http transport for self-hosted or cloud deployments. Once running, clients can connect through the MCP API to manage prompt lifecycles, browse workspaces and projects, inspect traces, and perform metrics and dataset operations. The server also provides resource templates and prompts to drive metadata-aware flows, enabling programmatic orchestration of Opik prompts and experiments. You can enable or disable specific toolsets to tailor capabilities to your use case (for example core, integration, or expert-dataset actions).
How to install
Prerequisites:
- Node.js v20.11.0 or newer
- npm (comes with Node.js)
Install and run from source:
git clone https://github.com/comet-ml/opik-mcp.git
cd opik-mcp
npm install
npm run build
Optional local configuration:
cp .env.example .env
Start options:
- Local stdio transport (for integration on the same machine as the client):
npm run start:stdio
- Remote HTTP transport (streamable-http):
npm run start:http
If you prefer running via npx directly (as shown in the Quickstart):
npx -y opik-mcp --apiKey YOUR_API_KEY
Additional notes
Environment and configuration tips:
- In remote (streamable-http) mode, authentication is enforced by default. By default the server expects Authorization: Bearer <OPIK_API_KEY> or x-api-key header values.
- Use STREAMABLE_HTTP_REQUIRE_AUTH and related STREAMABLE_HTTP_* env vars to customize remote behavior. See docs/streamable-http-transport.md for details.
- The toolsets system lets you limit capabilities (core, integration, expert-prompts, expert-datasets, expert-trace-actions, expert-project-actions, metrics, all). Configure with OPIK_TOOLSETS or CLI flags as described in docs/configuration.md.
- For local development, you can copy .env.example to .env to customize behavior without affecting production settings.
- If you run with npx, ensure you provide a valid API key or appropriate local auth configuration when connecting from a client.
- The repository ships the MCP server implementation only; there is no hosted remote MCP service in this repo by default.
Related MCP Servers
mcp-graphql
Model Context Protocol server for GraphQL
anytype
An MCP server enabling AI assistants to interact with Anytype - your encrypted, local and collaborative wiki - to organize objects, lists, and more through natural language.
one-search
š OneSearch MCP Server: Web Search & Scraper & Extract, Support agent-browser, SearXNG, Tavily, DuckDuckGo, Bing, etc.
kollektiv
Kollektiv MCP enables you to chat with and query your own documents directly from IDEs and MCP clients. Private, secure, and integrated into your favorite code editor
aica
aica(AI Code Analyzer) reviews your code using AI. Supports CLI and GitHub Actions.
oura
Oura Ring Model Controller Protocol (MCP).