能力标签

🔌 MCP 🤖 Agent 🔄 工作流 🐳 Docker 💻 CLI 🔗 REST API 🧬 Embedding 🎙 STT 🧠 Claude ✨ GPT

🔌

MCP工具

终端AI工具

基于 TypeScript · 让 AI 助手直接操作你的系统与工具

英文名：termpolis

⭐ 6 Stars 💻 TypeScript 📄 NOASSERTION 🏷 AI 8.0分

8.0AI 综合评分

ai-nativeclicross-platformtypescript

⚙️ 配置说明 📺 TG 频道

✦ AI Skill Hub 推荐

终端AI工具是 AI Skill Hub 本期精选MCP工具之一。综合评分 8.0 分，整体质量较高。我们强烈推荐将其纳入你的 AI 工具库，帮助提升工作效率。

📚 深度解析

终端AI工具是一款基于 MCP（Model Context Protocol）标准协议的 AI 工具扩展。MCP 协议由 Anthropic 开发并开源，旨在建立 AI 模型与外部工具之间的标准化通信接口，目前已被 Claude Desktop、Claude Code、Cursor 等主流 AI 工具采纳。

通过安装终端AI工具，你的 AI 助手将获得额外的工具调用能力，可以用自然语言直接操控该工具的功能，无需学习复杂的命令行语法。MCP 工具的核心价值在于"一次配置，永久增强"——配置完成后，每次与 AI 对话时都可以无缝调用这些工具。

在技术实现上，MCP 工具通过标准的 JSON-RPC 协议与 AI 客户端通信，工具的功能以"工具列表"的形式暴露给 AI 模型，AI 可以按需调用。终端AI工具提供了结构化的工具调用接口，使 AI 模型能够精确地理解和使用每个功能点，显著降低 AI 在工具使用上的错误率。

与传统的 API 集成相比，MCP 工具的优势在于无需编写代码——用户只需在配置文件中添加几行 JSON，即可让 AI 获得全新能力。AI Skill Hub 将终端AI工具评为 AI 评分 8.0 分，属于同类工具中的优质选择。

📋 工具概览

终端AI工具是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。

GitHub Stars

⭐ 6

开发语言

TypeScript

支持平台

Windows / macOS / Linux

维护状态

轻量级项目，按需更新

开源协议

NOASSERTION

AI 综合评分

8.0 分

工具类型

MCP工具

Forks

—

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

📌 核心特色

通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
提供结构化工具调用接口，显著降低 AI 集成复杂度
支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
可与其他 MCP 工具组合叠加，构建完整 AI 工作站
轻量无侵入设计，不影响现有系统架构

🎯 主要使用场景

在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
将多个 MCP 工具组合使用，构建个人专属 AI 工作站

以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。

安装命令

# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/codedev-david/termpolis

# 方式二：手动配置 claude_desktop_config.json
{
  "mcpServers": {
    "--ai--": {
      "command": "npx",
      "args": ["-y", "termpolis"]
    }
  }
}

# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json

📋 安装步骤说明

确认已安装 Node.js（v18 或以上版本）
打开 Claude Desktop 或 Claude Code 的 MCP 配置文件
按「交给 Agent 安装 → Claude Desktop」标签中的 JSON 配置填入 mcpServers 字段
保存配置文件并重启 Claude 客户端
重启后，在对话中即可使用本工具

以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。

常用命令 / 代码示例

# 安装后在 Claude 对话中直接使用
# 示例：
用户: 请帮我用 终端AI工具 执行以下任务...
Claude: [自动调用 终端AI工具 MCP 工具处理请求]

# 查看可用工具列表
# 在 Claude 中输入："列出所有可用的 MCP 工具"

以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。

配置示例

// claude_desktop_config.json 配置示例
{
  "mcpServers": {
    "__ai__": {
      "command": "npx",
      "args": ["-y", "termpolis"],
      "env": {
        // "API_KEY": "your-api-key-here"
      }
    }
  }
}

// 保存后重启 Claude Desktop 生效

📑 README 深度解析真实文档完整度 70/100 查看 GitHub 原文 →

以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

简介

Termpolis — Secure AI-Assisted Development

The open-source multi-agent terminal where Claude, Codex, Gemini, and Qwen work together as a team — without your source code leaving the machine.

🙏 We want to fix bugs fast. If anything is broken, please open an issue — we read every one.

Smart task routing assigns the right work to the right AI model. An intelligent conductor that knows who plays what instrument best. Local-first app — no Termpolis backend, no telemetry by default, no subscription. Free and open source — your agents, your terminal, your rules.

---

🎯 Flagship Feature: Multi-Agent Swarm with Smart Routing

Describe your task. Termpolis analyzes it, breaks it into subtasks, and assigns each to the AI model that's best at that type of work — automatically.

You: "Refactor the auth module, write comprehensive tests,
      document the API, and review for security vulnerabilities"

Termpolis Smart Router:
  📝 Refactor auth module    → Claude Code (100/100) — strongest at refactoring
  🧪 Write tests             → Codex (95/100)        — top-rated for test writing
  📄 Document the API        → Gemini CLI (90/100)    — best at docs, lowest cost
  🔒 Security review         → Claude Code (95/100)   — best at code review

  💰 Token Budget: Claude ~$0.60 | Codex ~$0.23 | Gemini ~$0.10 | Total ~$0.93

A dedicated Claude Code instance acts as the conductor — it reasons about your task, delegates subtasks to agents via MCP tools, and monitors completion. You watch it happen in the Swarm Dashboard. Every assignment is transparent and overridable. When the swarm finishes, a summary dialog shows what each agent accomplished.

Capability	Claude Code	Codex	Gemini CLI	Qwen Code
Refactoring	★★★★★	★★★★	★★★	★★★
Testing	★★★★	★★★★★	★★★	★★★
Documentation	★★★★	★★★★	★★★★★	★★★
Code Review	★★★★★	★★★	★★★★	★★★
DevOps/Infra	★★★	★★★	★★★★★	★★★
Bulk Tasks	★★★	★★★★	★★★	★★★★
Token Cost	$$$$	$$$	$$	$$

---

Features

AI-Native Features

AI Session Profiles — one-click launch profiles for Claude Code, Codex, Gemini CLI, and Qwen Code with custom profiles support
Command Palette — Ctrl+K opens a natural language command bar to control the app (new terminal, split panes, launch agents, run commands)
Prompt Templates — save reusable prompt snippets (Fix Tests, Code Review, Refactor, etc.) and insert them with Ctrl+Shift+P (accessible via Command Palette)
Multi-Agent Workflow Templates — built-in split-pane layouts (Claude + Shell, Full Stack Dev, Code Review) plus a visual editor to create, edit, and save your own custom workflows (name, icon, layout, 1–8 terminals each with shell + startup command + color, persisted across restarts)
Agent Status Detection — automatically detects when Claude Code, Codex, Gemini, or Qwen Code is running and shows a colored badge in the status bar
Cost Tracking — parses token usage and cost from AI agent output, displays running totals in the status bar
Session Recording — record terminal sessions with timestamps, export as shareable text logs
Output Pinning — pin important output blocks to a persistent panel that stays visible as the terminal scrolls
Diff Viewer — detects git diff output and renders it with syntax highlighting (green/red for additions/deletions)
Smart Context Panel — Ctrl+Shift+E opens a side panel showing file tree, git status, and recent commits for the current directory
Conversation History — Ctrl+Shift+I searches across all AI agent conversations indexed from terminal output
Voice Dictation — talk instead of type (Groq Whisper, opt-in, off by default): tap Ctrl+Shift+L to start/stop hands-free, or hold to talk; the hotkey and the send key are rebindable
Cost-Aware Model Picker — pin a Claude model per profile (launches with --model) or hot-swap a running agent's model live from the terminal header (/model), with savings hints — Sonnet ≈40% cheaper than Opus, Haiku ≈80% — so routine work runs cheap while hard work stays on Opus
Knowledge Graph — the shared memory stores typed connections between entries (bug → solved-by → fix, decision → supersedes → …), built explicitly via memory_link and automatically as curated memories are written. Agents follow the chain with memory_graph to reuse prior solutions fast instead of re-deriving them — the graph gets denser, and the agents get smarter, the more you use it
No duplicate data — every write is content-addressed (SHA-256 over normalized text); storing the same information twice is a no-op, so the vector store and the on-disk log never accumulate duplicates and never re-embed what they already hold

Bug Reports & Feature Requests

Found a bug or have an idea? Open an issue on GitHub:

Submit a Bug Report

Request a Feature

When reporting a bug, please include: - Your OS (Windows/macOS/Linux) and version - Termpolis version (shown in the title bar or package.json) - Steps to reproduce the issue - Expected vs actual behavior - Screenshots if applicable

Prerequisites

Node.js 18+
npm 9+
Windows only: Visual Studio Build Tools (for native node-pty compilation)
Linux only: build-essential, python3 (for native module compilation)

Installing the Linux .deb

Use dpkg, not sudo apt install ./termpolis*.deb. On Ubuntu 22.04+ apt drops to a sandboxed _apt user that can't read files in your home directory, which fails with "Permission denied / pkgAcquireRun: 13". dpkg doesn't drop privileges, so it works regardless of where the .deb is saved:

sudo dpkg -i ./termpolis_*.deb

That's the only command you need on v1.11.31+. The package's postinst takes care of the rest automatically:

runs apt-get install -f -y to pull any missing transitive deps (libgtk, libnss3, …),
refreshes the desktop + hicolor icon caches so the launcher icon shows up without a logout, and
the .desktop entry ships with --no-sandbox --disable-gpu baked into the Exec= line, so clicking the dock icon launches a working window on NVIDIA / Wayland setups where Chromium's GPU compositor would otherwise produce a blank black box.

If you ever need to launch from a shell with the same flags applied: /opt/Termpolis/termpolis --no-sandbox --disable-gpu.

**\ Windows SmartScreen note: SmartScreen may show a "Windows protected your PC" warning for newly signed software. Click "More info" then "Run anyway"* to proceed. Termpolis is digitally signed and safe to install — the warning disappears as download reputation builds.

Install & Run

git clone https://github.com/codedev-david/termpolis.git
cd termpolis
npm install
npm run dev

Building from Source

Windows (NSIS Installer)

bash scripts/download-tools.sh  # Download bundled CLI tools
npm run package

Output: dist-electron-builder/Termpolis Setup X.X.X.exe

Quick Start

CLI Tool

Control Termpolis from any terminal without the MCP protocol:

termpolis-cli health                    # Check if MCP server is running
termpolis-cli list                      # List all open terminals
termpolis-cli create "Dev" bash         # Create a new terminal
termpolis-cli run <id> "npm test"       # Run a command in a terminal
termpolis-cli read <id> 20              # Read last 20 lines of output
termpolis-cli close <id>                # Close a terminal
termpolis-cli files ~/projects          # List files at a path
termpolis-cli git ~/projects/myapp      # Get git status

MCP Server & Agent Integration

MCP Server — built-in HTTP/SSE server on localhost:9315 with 21 tools for AI agents to control terminals programmatically (incl. shared-memory search, write, list, the background primer, 1-hop memory_related traversal, and the knowledge graph: memory_link + multi-hop memory_graph)
Auto-registers with Claude Code — on launch, Termpolis injects itself into ~/.claude/settings.json so Claude Code can use it as an MCP server immediately. Zero configuration needed.
Stdio Adapter — for agents that use stdio-based MCP, a standalone adapter script proxies to the HTTP server
CLI Tool — termpolis-cli lets you control Termpolis from any terminal (list, create, run, read, close, files, git)
Auth Token — 256-bit random token per launch, required on all endpoints. Localhost only, CORS restricted.

Claude Code Integration

Termpolis auto-registers with Claude Code on launch — it adds itself to ~/.claude/settings.json so Claude Code sees it as an MCP server immediately. No manual configuration needed.

No Plugin System — By Design

Termpolis intentionally does not have a plugin or extension system
Third-party plugins are a major attack surface — they run with full app permissions, can access terminals, read output, and execute commands
Every feature in Termpolis is built-in, auditable, and ships with the app
If you need custom behavior, fork the repo — the codebase is open source and well-documented

🎯 aiskill88 AI 点评 A 级 2026-07-01

功能丰富的MCP工具，支持多窗口和集群