1. Conduid
  2. Developer Tools
  3. Mcpx Eval
MCP server · Developer Tools

Mcpx Eval

An open-ended eval framework for mcp.run tools

Unclaimed BSD-3-Clause last commit a year ago devtools
54Fair

Scored 2 days ago · breakdown

About Mcpx Eval

Mcpx Eval is an MCP server published by dylibso in the Developer Tools category: an open-ended eval framework for mcp.run tools. It has been installed 0 times through Conduid.

The repository has 22 stars and 6 forks, with the last commit a year ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues (0) before depending on it in production.

Install

Install
npx mcpx-eval

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

Ask AI about Mcpx Eval

Powered by Claude · Grounded in docs

I know everything about Mcpx Eval. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ·README presentNot checked yet.
  • ·License declaredNot checked yet.
  • ·Tests presentNot checked yet.
  • ·Dependencies pinnedNot checked yet.
  • ·No dynamic code executionNot checked yet.
  • !Scoped permissionsDoesn't declare a permission scope. Assume it can do anything its process can.

Releases

v0.4.3v0.4.3 · 21 May 2025Full Changelog**: https://github.com/dylibso/mcpx-eval/compare/v0.4.2...v0.4.3
v0.4.2v0.4.2 · 21 May 2025Full Changelog**: https://github.com/dylibso/mcpx-eval/compare/v0.4.1...v0.4.2
v0.4.1v0.4.1 · 12 May 2025Full Changelog**: https://github.com/dylibso/mcpx-eval/compare/v0.4.0...v0.4.1
v0.4.0v0.4.0 · 8 May 2025What's Changed refactor: use MCP protocol to access mcp.run via SSE/stdio by @zshipko in https://github.com/dylibso/mcpx-eval/pull/14 Full Changelog**: https://github.com/dylibso/mcpx-eval/compare/v0.3.0...v0.4.0
v0.3.0v0.3.0 · 16 Apr 2025What's Changed cleanup: rename ignore-tools -> ignored-tools by @zshipko in https://github.com/dylibso/mcpx-eval/pull/10 docs: add mcp.run setup by @zshipko in https://github.com/dylibso/mcpx-eval/pull/12 chore(docs): small typo by…

README

mcpx-eval

A framework for evaluating open-ended tool use across various large language models.

mcpx-eval can be used to compare the output of different LLMs with the same prompt for a given task using mcp.run tools. This means we're not only interested in the quality of the output, but also curious about the helpfulness of various models when presented with real world tools.

Test configs

The tests/ directory contains pre-defined evals

Installation

uv tool install mcpx-eval

Or from git:

uv tool install git+https://github.com/dylibso/mcpx-eval

Or using uvx without installation:

uvx mcpx-eval

mcp.run Setup

You will need to get an mcp.run session ID by running:

npx --yes -p @dylibso/mcpx gen-session --write

This will generate a new session and write the session ID to a configuration file that can be used by mcpx-eval.

If you need to store the session ID in an environment variable you can run gen-session without the --write flag:

npx --yes -p @dylibso/mcpx gen-session

which should output something like:

Login successful!
Session: kabA7w6qH58H7kKOQ5su4v3bX_CeFn4k.Y4l/s/9dQwkjv9r8t/xZFjsn2fkLzf+tkve89P1vKhQ

Then set the MCP_RUN_SESSION_ID environment variable:

$ export MCP_RUN_SESSION_ID=kabA7w6qH58H7kKOQ5su4v3bX_CeFn4k.Y4l/s/9dQwkjv9r8t/xZFjsn2fkLzf+tkve89P1vKhQ

Usage

Run an eval comparing all mcp.task runs for my-task:

mcpx-eval test --task my-task --task-run all

Only evaluate the latest task run:

mcpx-eval test --task my-task --task-run latest

Or trigger a new task run:

mcpx-eval test --task my-task --task-run new

Run an mcp.run task locally with a different set of models:

mcpx-eval test --model .. --model .. --task my-task --iter 10

Generate an HTML scoreboard for all evals:

mcpx-eval gen --html results.html --show

Test file

A test file is a TOML file containing the following fields:

  • name - name of the test
  • task - optional, the name of the mcp.run task to use
  • task-run - optional, one of latest, new, all or the name/index of the task run to analyze
  • prompt - prompt to test, this is passed to the LLM under test, this can be left blank if task is set
  • check - prompt for the judge, this is used to determine the quality of the test output
  • expected-tools - list of tool names that might be used
  • ignored-tools - optional, list of tools to ignore, they will not be available to the LLM
  • import - optional, includes fields from another test TOML file
  • vars - optional, a dict of variables that will be used to format the prompt

README mirrored from the source repository 2 days ago. The original is authoritative.

Questions

About Mcpx Eval

How do I install Mcpx Eval?

Run npx mcpx-eval, then add the server to your MCP client's configuration. Conduid has recorded 0 installs, so the command is known to work with current clients.

Is Mcpx Eval safe to use with an AI agent?

Its trust score is 54 out of 100 (fair). It passes 0 of 1 static security checks; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Mcpx Eval still maintained?

The last commit was a year ago, with 0 open issues. That's long enough that you should check whether the maintainer is responding to issues before depending on it.