LABSCO SUMMARY

The prompts your team versions in MLflow, usable in the desktop client.

Templates stay where they are governed and get compiled on the way out with the variables you supply, so a repetitive workflow runs from the reviewed version rather than from a copy someone pasted into a chat months ago.

What it is

A server for MLflow Prompt Registry that lists the prompts stored there and compiles one with the arguments you pass.

What you get

The prompts available in the registry, with a filter and cursor-based pagination (list-prompts)
One prompt retrieved and compiled against a JSON object of variables (get-prompt)
Prompts surfaced through MCP's prompts interface, so they appear in the client for repetitive tasks and common workflows

Requirements

An MLflow server hosting the Prompt Registry — a recent mlflow, started with mlflow server — and its address in the environment. Built from source with npm install and npm run build.

Setup effort

Follow the published steps — pip install mlflow>=2.21.1

See installation options

USE CASES

01

Running a team's approved prompt instead of retyping it

Find it in the registry and pull the compiled version into the conversation.

list-promptsget-prompt

02

Filling a template with today's inputs

Pass the variables as a JSON object and get the compiled prompt back.

get-prompt

03

Finding out which prompts the registry actually holds

List them with a filter, paging through if there are many.

list-prompts

04

Keeping a recurring task on the version that was reviewed

Fetch from the registry each run, so a change made there takes effect everywhere.

get-prompt

2 tools

Listed in project docs. Your client may show a different list.

list-prompts

get-prompt

BEFORE YOU INSTALL

What to know first

Tools you get

2 tools named in the project's documentation. list-prompts and get-prompt serve prompt templates, so a client may show them as prompts rather than callable tools.

Where it runs

The documented client runs the server on your machine.

INSTALLATION

Connect your client

Server setup

TERMINAL

pip install mlflow>=2.21.1
mlflow server --port 5000
npm install
npm run build

MCP Clients

Claude Desktop

CLAUDE DESKTOP · JSON

{
  "mcpServers": {
    "mlflow": {
      "command": "node",
      "args": [
        "<YOUR_ABSOLUTE_PATH_TO_REPOSITORY>/dist/index.js"
      ],
      "env": {
        "MLFLOW_TRACKING_URI": "http://localhost:5000"
      }
    }
  }
}

MLflow Prompt Registry