Terminal Coding Agent

中文文档 | English

A powerful CLI-based multi-agent coding assistant powered by Claude Agent SDK.

Rust version available: rust-terminal-coding-agent

Features

• Multi-Agent Architecture - Coordinator, Reader, Coder, Reviewer agents working together
• Smart Routing - Automatically routes tasks to the most suitable agent
• Skills System - Extensible skills for specialized tasks (code review, git commit, PDF analysis, etc.)
• Interactive UI - / command menu and @ file browser for easy interaction
• Web Scraping - Built-in Playwright integration for web content extraction
• Deep Research - Dify-powered comprehensive research workflow

Skills

Built-in skills that extend the agent's capabilities:

Skill	Description	Agent
/code-review	Analyze code quality, find bugs, security issues	Reviewer
/git-commit	Create well-structured conventional commits	Coder
/pdf-analyze	Extract text/tables from PDFs, fill forms	Reader
/web-scrape	Scrape web content with Playwright (internal)	Coordinator
/deep-research	Comprehensive research via Dify workflow (internal)	Coordinator

Create Custom Skills

Create skills in .claude/skills/<skill-name>/SKILL.md:

---
name: my-skill
description: Description of what this skill does
version: 1.0.0
allowed-tools:
  - Read
  - Write
  - Bash
---

# My Custom Skill

Instructions for Claude on how to execute this skill...

Quick Start

1. Clone and Install

git clone https://github.com/yourusername/terminalAgent.git
cd terminalAgent
npm run install:all

2. Configure Environment

Copy the example environment file and add your API key:

cp deepresearch/.env.example deepresearch/.env

Edit deepresearch/.env:

ANTHROPIC_API_KEY=your-api-key-here
ANTHROPIC_BASE_URL=https://api.anthropic.com

Supports custom API endpoints (e.g., API proxy services).

3. Run

# Development mode
npm run dev

# Or build and run
npm run build
npm start

4. Global CLI (Optional)

cd deepresearch
npm link

# Now you can use from anywhere:
agent                    # Interactive mode
agent "your question"    # Single query
agent /code-review       # Invoke a skill

Usage

Commands

Command	Description
/help	Show help
/mode	Switch permission mode (safe/unsafe)
/clear	Start new session
/exit	Exit program
@	Open file browser
@file.ts	Attach file to context

Architecture

┌─────────────────────────────────────────────────────────────┐
│                    User Input                                │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│                  Smart Router                                │
│         (Keyword matching, Skill detection)                  │
└───────┬─────────┬─────────┬─────────┬───────────────────────┘
        │         │         │         │
        ▼         ▼         ▼         ▼
    ┌───────┐ ┌───────┐ ┌───────┐ ┌───────────┐
    │Reader │ │ Coder │ │Review │ │Coordinator│
    │  📖   │ │  💻   │ │  🔍   │ │    🎯     │
    └───────┘ └───────┘ └───────┘ └───────────┘
        │         │         │         │
        └─────────┴─────────┴─────────┘
                          │
                          ▼
              ┌───────────────────────┐
              │   [DISPATCH:agent]    │
              │   Inter-agent calls   │
              └───────────────────────┘

Agent Types

Agent	Icon	Role
Coordinator	🎯	Task decomposition, multi-agent dispatch
Reader	📖	Code reading, analysis, understanding
Coder	💻	Code writing, modification, implementation
Reviewer	🔍	Code review, quality checks, bug detection

Available Tools

Tool	Description
Read	Read file contents
Write	Create/overwrite files
Edit	Edit existing files
Bash	Execute shell commands
Glob	Find files by pattern
Grep	Search text in files
LSP	Language Server Protocol integration
WebFetch	Fetch web content
WebSearch	Search the web
Skill	Invoke skills

Deep Research (Dify Integration)

This project integrates Dify workflows for deep research capabilities.

Workflow Architecture

Start → Background Search → Task Analysis → Research Loop → Implementation Guide → End
                                 ↓
                      [DeepSeek Reasoner]
                                 ↓
              ┌──────────────────────────┐
              │   Research Loop (1-5x)   │
              │  ┌─────────────────────┐ │
              │  │ Plan → Search → Reason│ │
              │  └─────────────────────┘ │
              └──────────────────────────┘

Models Used

Node	Model	Purpose
Task Analysis	DeepSeek Reasoner	Deep analysis of coding tasks
Query Planner	GPT-5	Generate precise search queries
Code Reasoning	DeepSeek Reasoner	Extract actionable code info
Implementation Guide	GPT-5	Generate complete implementation

Configure Dify

1. Create an account at Dify
2. Import the dify/code-research-skill.yml workflow
3. Configure in .env:

DIFY_API_KEY=your-dify-api-key
DIFY_BASE_URL=https://api.dify.ai/v1

Project Structure

terminalAgent/
├── deepresearch/
│   ├── src/
│   │   ├── index.ts              # Main entry point
│   │   ├── agents/               # Agent implementations
│   │   │   ├── base.ts           # BaseAgent class
│   │   │   ├── coordinator.ts    # Coordinator agent
│   │   │   ├── reader.ts         # Reader agent
│   │   │   ├── coder.ts          # Coder agent
│   │   │   └── reviewer.ts       # Reviewer agent
│   │   ├── core/
│   │   │   ├── router.ts         # Smart routing logic
│   │   │   └── session.ts        # Session management
│   │   ├── config/
│   │   │   ├── agents.ts         # Agent configurations
│   │   │   └── constants.ts      # Constants
│   │   ├── ui/
│   │   │   ├── smart-input.ts    # Input with "/" and "@"
│   │   │   ├── commands.ts       # Command picker
│   │   │   └── file-browser.ts   # File browser
│   │   └── prompts/              # Agent system prompts
│   ├── .claude/
│   │   └── skills/               # Skill definitions
│   ├── bin/agent.cjs             # CLI entry point
│   └── package.json
├── dify/                         # Dify workflow configs
│   └── code-research-skill.yml   # Deep research workflow DSL
├── docs/images/                  # Documentation images
├── CLAUDE.md                     # Instructions for Claude Code
└── README.md

Environment Variables

Variable	Description	Required
ANTHROPIC_API_KEY	Your Anthropic API key	Yes
ANTHROPIC_BASE_URL	API endpoint (default: api.anthropic.com)	No
ANTHROPIC_MODEL	Model to use (default: claude-sonnet-4-20250514)	No
DIFY_API_KEY	Dify API key for deep-research skill	No
DIFY_BASE_URL	Dify API endpoint	No

Tech Stack

• TypeScript - Type-safe development
• @anthropic-ai/claude-agent-sdk - Claude Agent SDK for multi-agent orchestration
• Playwright - Web automation and scraping
• Cheerio - HTML parsing and manipulation
• Dify - Deep research workflow engine
• dotenv - Environment configuration

Development

# Install dependencies
npm run install:all

# Development mode with hot reload
npm run dev

# Build TypeScript
npm run build

# Run compiled version
npm start

# Test UI components
cd deepresearch && npm run test:ui

License

MIT

Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request