---
title: "Manage artifacts via AI agents and LLMs using Cloudsmith MCP"
description: "Cloudsmith MCP lets you connect AI agents and large language models (LLMs) like ChatGPT and Claude directly to your software supply chain"
canonical_url: "https://cloudsmith.com/product/mcp-server"
last_updated: "2026-02-17T09:16:29Z"
---
# Manage artifacts via AI agents and LLMs using Cloudsmith MCP

Cloudsmith MCP server

## Control your software supply chain from your AI agent using Cloudsmith MCP

Model Context Protocol (MCP) servers are software bridges between AI models and data sources like Cloudsmith.  Using the Cloudsmith MCP server, your LLMs and AI agents can interact with artifacts and data in your Cloudsmith workspace, and gain insights into your software supply chain. 

[Book a demo](/book-a-demo)

Use Cases

## What can you do with Cloudsmith MCP? 
**We can’t wait to show you.**

With Cloudsmith MCP you can ask questions about how software artifacts like packages, containers, and libraries are flowing through your software supply chain. Query for usage metrics. Even control and manage artifacts and policies.  Ultimately, virtually anything that can be done with the Cloudsmith API may be accessible via MCP. 

[Why we're working on MCP at Cloudsmith](/blog/prototyping-an-mcp-server)

### Bring your own LLM

Securely connect your Cloudsmith workspace to Anthropic’s Claude, ChatGPT Pro, or your LLM of choice.

### Monitor usage & gain insight

Ask Cloudsmith questions like “what’s the most utilized package in my Cloudsmith workspace?” or “is Docker usage trending up or down this month compared to last month?” or “which users are consuming the most data this month?” If you can access a metric from the Cloudsmith API, you’ll be able to use it in context from your AI agent.

### Define software policies

Manage your policies from your AI agent. You might ask Cloudsmith to “create a policy that blocks all python packages younger than a week old” or “create a policy that alerts us if any critical vulnerabilities are found”. (For now, we’re not allowing activation of policies via MCP; that’ll remain under manual control.)

### Perform actions on your repositories, packages, teams and users

Take actions directly using LLMs. For now, we’re only allowing non-destructive actions; over time we’ll expose more APIs.