Opik MCP Server
Wire your coding agent to Opik so you can inspect LLM traces, prompts, datasets, and metrics without leaving the editor.
Overview
Opik MCP Server is a MCP server for the Ship phase that lets coding agents read and work with Opik prompts, traces, datasets, and metrics over stdio.
What is this MCP server?
- stdio MCP server (npm opik-mcp v2.0.2) runnable via npx with -y for non-interactive install
- Tools cover Opik prompts, traces, datasets, and workspace metrics over the official SDK
- Requires OPIK_API_KEY (secret) for authenticated Opik workspace calls
- Model Context Protocol integration for Claude Code, Cursor, and other MCP clients—not a standalone skill file
- Published server version 2.0.2
- npm package identifier opik-mcp with stdio transport
- OPIK_API_KEY required secret environment variable
Community signal: 208 GitHub stars.
What problem does it solve?
Debugging agent quality in a separate Opik tab breaks flow and makes it hard to tie a bad trace to the code change you are about to ship.
Who is it for?
Solo builders already on Opik who ship Claude Code or Cursor agents and want MCP-native access to eval artifacts during test and pre-release checks.
Skip if: Teams with no Opik workspace, no LLM tracing needs, or builders who only need generic log grep without structured LLM datasets.
What do I get? / Deliverables
After registration, your agent can pull Opik trace, prompt, dataset, and metric context inline so you can iterate on LLM behavior with evidence in the same thread.
- Registered stdio MCP server talking to Opik
- Agent-callable access to prompts, traces, datasets, and metrics
- Faster trace-to-code iteration during LLM ship checks
Recommended MCP Servers
Journey fit
LLM observability and evaluation belong in Ship because you use traces and metrics to validate agent behavior before and after releases. Testing is the canonical shelf for dataset-backed evals, prompt versions, and trace review against production-like runs.
How it compares
MCP bridge to Opik observability, not an in-repo test runner or a generic monitoring stack for non-LLM services.
Common Questions / FAQ
Who is Opik MCP Server for?
Indie and solo developers using Opik to monitor LLM apps who want their MCP-enabled agent to query prompts, traces, datasets, and metrics directly.
When should I use Opik MCP Server?
Use it during Ship when you are validating agent outputs, comparing prompt versions, or investigating regressions using Opik traces and datasets before release.
How do I add Opik MCP Server to my agent?
Add the npm package opik-mcp with stdio transport (typically npx -y opik-mcp), set OPIK_API_KEY in the server environment, and register the server in your client’s MCP config.