# mcpc — Universal MCP CLI Client

> A Swiss Army knife CLI for the Model Context Protocol: connect to MCP servers, manage sessions, list tools, and script tool calls via JSON output.

## Install

Copy the content below into your project:

## Quick Use

1. Install / run:
   ```bash
   npm install -g @apify/mcpc
   ```
2. Start / smoke test:
   ```bash
   mcpc --help
   ```
3. Verify:
   - Connect to one MCP server, run `tools-list`, then call a simple tool and confirm JSON mode works for scripting.

## Intro

A Swiss Army knife CLI for the Model Context Protocol: connect to MCP servers, manage sessions, list tools, and script tool calls via JSON output.

- **Best for:** Developers and agents who want a single CLI surface for MCP (interactive + scripting) instead of bespoke integrations
- **Works with:** Node.js CLI; HTTP/stdio MCP servers; shell scripting with jq/xargs (per README)
- **Setup time:** 8 minutes

## How to Use It Well

- Setup time ~8 minutes (npm install + connect one server)
- GitHub stars + forks (verified): see Source & Thanks
- Supports persistent named sessions (e.g., `@test`) for repeated tool calls

Use `mcpc` as the thin waist: everything agent-side is a Bash tool, everything server-side is MCP. That makes it easier to standardize sessions, auth, logging, and scripting patterns across teams.

### FAQ

**Q: Why a CLI for MCP?**
A: Agents already have a universal interface—shell access. A CLI lets agents use MCP without custom function wiring.

**Q: Is it interactive or scriptable?**
A: Both. It has an interactive shell mode and a JSON output mode for pipelines.

**Q: How do I keep credentials safe?**
A: Use OAuth profiles as described in the README and run MCP proxying/sandboxing if you share a machine with multiple agents.

## Source & Thanks

> Source: https://github.com/apify/mcpc
> License: Apache-2.0
> GitHub stars: 596 · forks: 56

---

<!-- ZH -->

## 快速使用

1. 安装 / 运行：
   ```bash
   npm install -g @apify/mcpc
   ```
2. 启动 / 冒烟测试：
   ```bash
   mcpc --help
   ```
3. 验证：
   - Connect to one MCP server, run `tools-list`, then call a simple tool and confirm JSON mode works for scripting.

## 简介

通用 MCP 命令行客户端：连接 HTTP/stdio MCP server、维护会话、列出工具并调用工具，支持 JSON 输出便于脚本化，也适合让 agent 通过 shell 使用 MCP。

- **适合谁：** 希望用一个 CLI 统一操作 MCP（交互 + 脚本），避免为每个 server 写专用集成的开发者与 agent
- **可搭配：** Node.js CLI；HTTP/stdio MCP server；配合 jq/xargs 做脚本化（以 README 为准）
- **准备时间：** 8 分钟

## 实战建议

- 接入约 8 分钟（npm 安装 + 连接一个 server）
- GitHub stars + forks（已核验）：见「来源与感谢」
- 支持持久化命名会话（例如 `@test`），便于重复调用工具

可以把 `mcpc` 当“细腰”：agent 侧只需要 Bash 工具，server 侧统一用 MCP。这样更容易在团队内标准化会话、鉴权、日志与脚本化模式。

### FAQ

**为什么要用 CLI 操作 MCP？**
A: 因为对 agent 来说最通用的接口就是 shell。CLI 能让 agent 用 MCP，而无需为每个 server 单独接函数。

**它是交互式还是脚本化？**
A: 两者都支持：有交互 shell，也支持 JSON 输出方便管道脚本。

**凭据怎么更安全？**
A: 按 README 说明使用 OAuth profile，并在多 agent 场景开启代理/隔离以保护凭据。

## 来源与感谢

> Source: https://github.com/apify/mcpc
> License: Apache-2.0
> GitHub stars: 596 · forks: 56


---
Source: https://tokrepo.com/en/workflows/mcpc-universal-mcp-cli-client
Author: MCP Hub