Integrations

MCP Server Support for Cloudsmith CLI

How to integrate MCP with Cloudsmith

The Cloudsmith CLI now includes support for the Model Context Protocol (MCP), allowing AI assistants like Claude to interact with your Cloudsmith repositories through natural language.

What's New

The MCP server enables AI assistants to programmatically manage Cloudsmith repositories, packages, and artifacts by automatically generating tools from Cloudsmith's OpenAPI specification.

Requirements

Python 3.10 or later is required to use the Cloudsmith CLI.

Getting Started

Auto-Configure MCP Clients

The CLI provides a command to facilitate the configuration of MCP clients:

bash

# Configure for default profile
$ cloudsmith mcp configure

# Configure for specific profile
$ cloudsmith mcp configure -P my-profile

Currently supported MCP clients for auto-configuration: Claude Desktop, Cursor, VS Code (Copilot), and Gemini CLI.

Manual Configuration

Alternatively, you can manually configure the MCP server by modifying the JSON configuration file for your LLM client.

For example, with Claude Desktop on macOS, the configuration file is located at:

~/Library/Application Support/Claude/claude_desktop_config.json

Example configuration:

json

{
  "mcpServers": {
    "cloudsmith": {
      "command": "cloudsmith",
      "args": [
        "mcp",
        "start"
      ]
    }
  }
}

Note: The MCP server will be started automatically by your LLM client. You don't need to run cloudsmith mcp start manually.

Managing Available Tools

Due to the large number of tools exposed by the MCP server, a limited subset is available by default to avoid filling the LLM's context window.

Disabled tool categories by default: broadcasts, rates, packages_upload, packages_validate, user_token, user_tokens, webhooks, status, repos_ecdsa, repos_geoip, repos_gpg, repos_rsa, repos_x509, repos_upstream, orgs_openid, orgs_saml, orgs_invites, files, badges, quota, users_profile, workspaces_policies, storage_regions, entitlements, metrics_entitlements, metrics_packages, orgs_teams, repo_retention

List Available Tools and Groups

bash

# Show all available tools
$ cloudsmith mcp list_tools -a

# Show all tool groups
$ cloudsmith mcp list_groups -a

After reviewing the available tools and groups, you can customize which ones are exposed to your AI assistant.

Customizing Available Tools

You can customize which tools are available by editing ~/.cloudsmith/config.ini:

bash

[default]
mcp_allowed_tools=workspaces_policies_simulate_list
mcp_allowed_tool_groups=metrics

This configuration exposes:

Individual tools listed in mcp_allowed_tools
All tools within the groups specified in mcp_allowed_tool_groups

Authentication

The MCP server uses your existing Cloudsmith CLI credentials:

API Keys: Automatically used if configured
SSO: Complete authentication via cloudsmith auth before configuring and launching your LLM client, as MCP clients cannot trigger the SSO flow automatically once the server is running

Multi-Profile Support

Configure separate MCP server instances for different Cloudsmith profiles using the -P flag with the configure command.

Updated 2 weeks ago