Overview

mcporter (MCPorter 🧳) is an open-source TypeScript runtime, CLI, and code-generation toolkit for the Model Context Protocol (MCP). It lets developers and AI agents discover configured MCP servers, call tools directly, compose automations in TypeScript, and generate single-purpose CLIs from any MCP endpoint.

Built to support Anthropic’s “Code Execution with MCP” guidance, it eliminates boilerplate and context-window bloat by letting AI agents invoke tools via lightweight CLI binaries instead of loading massive JSON schemas.

Latest version: 0.8.2 (March 2026). MIT licensed.

Features

Zero-Config Discovery: Automatically finds MCP servers from ~/.mcporter/mcporter.json, project configs, and integrations with Claude, Cursor, Codex, VS Code, Windsurf, and OpenCode.
CLI-First Tool Calling: Run tools with natural syntax (mcporter call linear.create_comment ...) or function-call style — no schema knowledge needed.
One-Command CLI Generation: mcporter generate-cli creates standalone, self-contained CLIs (Bun/Node) from any MCP server for easy sharing and agent use.
Typed TypeScript API: mcporter emit-ts generates .d.ts interfaces or full client wrappers with strong typing, defaults, and validation.
Composable Runtime: createRuntime() + createServerProxy() with connection pooling, OAuth caching, log tailing, and support for HTTP/SSE/stdio transports.
Ad-Hoc & Daemon Mode: Connect to any MCP endpoint instantly; keep stateful servers (e.g. chrome-devtools) warm with daemon commands.
AI-Agent Friendly: Reduces context bloat — agents call a tiny CLI binary instead of embedding full MCP schemas.
Robust Tooling: Auto-correction, rich output formatting (text(), json()), config management, and debug helpers.

Use Cases

AI Agent Tool Integration: Let Claude/Cursor call complex MCP servers via generated CLIs without exploding context windows.
Rapid Automation Scripting: Build TypeScript workflows that compose multiple MCP tools with full type safety.
Tool Sharing: Generate and distribute single-purpose CLIs (e.g. linear-comment) for teams or public use.
Development & Debugging: Instantly inspect any MCP server schema or call tools ad-hoc during prototyping.
Reducing MCP Bloat: Replace heavy schema loading in editors/agents with lightweight CLI calls.

Installation & Quick Start

Run instantly (no install)

npx mcporter list
npx mcporter list linear --schema

Global install

npm install -g mcporter
# or Homebrew
brew tap steipete/tap && brew install steipete/tap/mcporter

Example calls

# Colon syntax
mcporter call linear.create_comment issueId:ENG-123 body:'Looks good!'

# Function-call syntax
mcporter call 'linear.create_comment(issueId: "ENG-123", body: "Looks good!")'

# Generate a CLI
mcporter generate-cli --command https://mcp.context7.com/mcp

TypeScript usage

import { createRuntime } from "mcporter";
const runtime = await createRuntime();
const result = await runtime.callTool("context7", "resolve-library-id", {
  args: { libraryName: "react" }
});
console.log(result.text());
await runtime.close();

Full docs, CLI reference, and examples in the repository.

Why mcporter?

MCP gives AI agents superpowers through tools, but managing schemas and connections is painful. mcporter makes MCP ergonomic — discover once, call instantly, generate CLIs for agents, and write clean typed code. The ultimate MCP Swiss Army knife.

mcporter