Skip to main content
mage-agent exposes a Model Context Protocol (MCP) server over stdio. Any MCP-compatible AI client can connect by launching mage-agent mcp as a subprocess.

​
Prerequisites

For manual MCP configuration, complete Installation and Authentication before setting up IDE integration. For Claude Desktop, use the bundled Desktop extension and its setup flow. It includes the executable and does not require the Claude Code CLI or a Python installation.

​
Cursor

1

Open MCP settings

In Cursor, go to Settings → MCP (or open the Command Palette and search for “MCP Servers”).
2

Add mage-agent

Add the following server config:
3

Reload Cursor

Save the config and reload the window. Confirm that mage-agent tools appear in the MCP tool list.

​
Claude Code

1

Add the MCP server via CLI

Or add it manually to .claude/settings.json in your project (or ~/.claude/settings.json globally):
2

Restart Claude Code

Restart the Claude Code session. Run a quick sanity check:
“Use the pipeline_list tool to list my Mage pipelines”

​
OpenAI Codex CLI

1

Edit your Codex config

Add mage-agent to the mcpServers section of your Codex config file:
2

Restart Codex CLI

Restart Codex CLI and verify tools and resources are visible.

​
PATH issues

If mage-agent is not found when the IDE launches the MCP server, use the absolute path to the executable:

​
Skill resources

When connected, your AI assistant can read bundled Markdown guidance through MCP resources. The catalog includes pipeline, execution, data preview, and chart guidance. Examples: