skforecast-ai-mcp

The MCP server of skforecast-ai: deterministic time series forecasting as tools for coding agents (Claude Code, Cursor, VS Code, Codex, Claude Desktop and other MCP clients).

The agent brings the language model and calls skforecast-ai as tools: profile a CSV file, plan a forecaster, backtest it, compare candidates and forecast. Every decision comes from rule-based code, every result comes with the script that produced it, and responses never carry rows of your data.

Usage

The server runs over stdio and needs one directory, --allow-dir: it only reads CSV files inside it. With uv:

uvx skforecast-ai-mcp --allow-dir /absolute/path/to/project

As the entry of an MCP client:

{
  "mcpServers": {
    "skforecast-ai": {
      "command": "uvx",
      "args": ["skforecast-ai-mcp", "--allow-dir", "/absolute/path/to/project"]
    }
  }
}

The setup of each client, the tools, every option and what the server does with your data are in MCP server for coding agents.

What this package is

A launcher with no logic of its own. It installs skforecast-ai[mcp] at the same version and runs skforecast-ai mcp with the arguments it receives, so these two commands start the same server:

uvx skforecast-ai-mcp --allow-dir /absolute/path/to/project
uvx --from "skforecast-ai[mcp]" skforecast-ai mcp --allow-dir /absolute/path/to/project

It exists because the MCP registry names a package, not one of its extras.

License

Apache-2.0, as skforecast-ai.