ToonCode — 🇹🇭 Thai Coding Agent CLI

by VotaLab | v2.6.1 | Thai AI Coding Agent powered by free AI models

  ████████╗ ██████╗  ██████╗ ███╗   ██╗ ██████╗ ██████╗ ██████╗ ███████╗
     ██║   ██║   ██║██║   ██║██╔██╗ ██║██║     ██║   ██║██║  ██║█████╗
     ╚═╝    ╚═════╝  ╚═════╝ ╚═╝  ╚═══╝ ╚═════╝ ╚═════╝ ╚═════╝ ╚══════╝

ToonCode is a free, open-source Thai AI coding agent that runs in your terminal. It uses free AI models (no API key needed) and provides a professional coding experience with 21 tools, 32 skills, multi-agent teams, semantic search, browser automation, MCP server support, and more.

Features

21 Built-in Tools

ToonCode comes with a comprehensive set of tools that the AI can use autonomously:

| Category | Tool | Description | |----------|------|-------------| | File | read | Read files with line numbers | | | write | Create/overwrite files (shows diff) | | | edit | String replacement in files (shows diff) | | | multi_edit | Batch edits across multiple files in one call | | | glob | Find files by pattern (e.g. **/*.py) | | | grep | Search file contents with regex | | | list_dir | Browse directory structure | | Shell | bash | Run shell commands (auto-detects PowerShell on Windows, bash on Mac/Linux) | | Web | web_search | Search the web via DuckDuckGo | | | web_fetch | Fetch and extract web page content | | | http | Full HTTP client — GET/POST/PUT/DELETE with headers, auth, JSON body | | Browser | browser | Playwright browser automation — open pages, click, fill forms, take screenshots, read console, monitor network | | System | screenshot | Capture desktop or window screenshot with OCR | | Agent | spawn_agent | Spawn sub-agents (coder, reviewer, tester, researcher) | | | bosshelp | Escalate difficult tasks to Boss AI (more powerful model) | | Memory | memory_save | Save notes for future sessions | | | memory_search | Search past memories by keyword | | Tasks | task_create | Create a tracked task | | | task_update | Update task status (pending/in_progress/done) | | | task_list | View all tasks and progress |

32 Built-in Skills

Skills are pre-built prompt templates organized by category. Use them with slash commands:

| Category | Skills | Example | |----------|--------|---------| | Code (10) | refactor, review, test, lint, explain, optimize, scaffold, types, convert, deps | /refactor src/utils.py | | Debug (4) | fix, error, perf, trace | /fix — auto-detect and fix bugs | | Git (3) | commit, changelog, pr | /commit — smart commit message | | DevOps (2) | docker, ci | /docker — generate Dockerfile | | Docs (2) | readme, api | /readme — generate README | | Data (3) | migrate, seed, sql | /sql — query helper | | API (2) | endpoint, mock | /endpoint — create new API endpoint | | Architecture (3) | diagram, pattern, split | /diagram — generate architecture diagram | | Security (2) | audit, secrets | /audit — OWASP Top 10 security audit |

Slash Commands

| Command | Description | |---------|-------------| | /help | Show all commands | | /model [name] | Switch AI model | | /boss <task> | Create detailed implementation plan using Boss AI | | /team <task> | Start multi-agent team (see below) | | /plan | Toggle Plan Mode (AI plans only, no edits) | | /do | Execute pending tasks | | /tasks | Show task progress | | /continue, /c | Continue where AI left off | | /save, /s | Save session | | /resume, /r | Load saved session | | /compact | Compress conversation context | | /memory | Browse saved memories | | /semantic <query> | Semantic search across codebase (requires LanceDB) | | /index | Rebuild codebase structure index | | /paste, /v | Send clipboard content to AI | | /send <msg> | Chat between ToonCode windows | | /bg, /ps | Manage background processes | | /commit | Git add & commit with smart message | | /diff | Git diff | | /status | Git status | | /undo | Undo last file edit | | /init | Create TOONCODE.md (auto-imports CLAUDE.md etc.) | | /config | Show/edit configuration | | /skills | List all 32 built-in skills | | /cost | Token usage statistics |

Multi-Agent Team System

ToonCode's most powerful feature. Spawn a team of specialized AI agents that coordinate and collaborate on complex tasks:

/team สร้างเว็บ portfolio             # Default: planner + frontend + backend + reviewer
/team build REST API --roles planner,backend,tester   # Custom roles

13 Agent Roles

| Role | Emoji | Thai Name | Specialty | |------|-------|-----------|-----------| | Planner | 📋 | สมชาย | Breaks tasks into steps, assigns work to agents | | Supervisor | 🧠 | สุภาพร | Monitors progress, resolves conflicts | | Frontend | 🎨 | ณัฐพล | Builds UI — HTML/CSS/JS/React | | Backend | ⚙️ | วิชัย | APIs, databases, server logic | | Reviewer | 🔍 | มาลี | Code review, finds bugs, suggests improvements | | Tester | 🧪 | ธนา | Writes and runs tests | | Researcher | 🔬 | ปิยะ | Finds info, gathers data, provides sources | | Analyst | 📊 | นภา | Data analysis, comparisons, insights | | Writer | ✍️ | กานดา | Articles, docs, reports, plans | | Editor | 📝 | อรุณ | Reviews and improves written content | | Designer | 🎯 | พิมพ์ | Architecture, schemas, UI layouts |

How Teams Work

1. Planner receives the task and breaks it into sub-tasks
2. Tasks are assigned to specialized agents via JSON messages on a shared channel
3. Each agent works independently with access to all 21 tools
4. A shared task board tracks progress (pending / in_progress / done)
5. Agents communicate results back through the shared channel
6. Supervisor monitors progress and resolves conflicts when needed
7. Agents can dynamically spawn new agents if the task requires it

Available AI Models

| Model | Context Window | Notes | |-------|---------------|-------| | big-pickle | 200K | Default — strong at coding tasks | | minimax-m2.5-free | 204K | Good all-round performance | | nemotron-3-super-free | 131K | Fast execution | | gpt-5-nano | 1M | Largest context for big projects |

• Auto-fallback: if a model fails, ToonCode automatically tries the next available model
• Per-model config: each model can have its own API URL and key
• Switch models anytime with /model <name>

Project Context Awareness

ToonCode automatically reads and understands your project:

• Context Files — auto-imports TOONCODE.md, CLAUDE.md, GEMINI.md, .cursorrules, COPILOT.md
• Codebase Indexing — scans project structure, extracts function/class symbols from code
• Tech Stack Detection — auto-detects frameworks (FastAPI, React, Next.js, Django, Express, Go, Rust, etc.)
• Semantic Search — optional vector-based code search using LanceDB for finding similar patterns

Browser Automation

Control a real browser powered by Playwright:

❯ เปิด google ค้นหา "ToonCode" แล้วสรุปผลลัพธ์

AI → browser open https://google.com/search?q=ToonCode
AI → browser text        # reads page content
AI → browser screenshot  # captures the page
AI → browser console     # checks JS errors
AI → browser network     # monitors API calls

Requires: pip install playwright && playwright install chromium

MCP Server Support

Extend ToonCode with custom tools via Model Context Protocol servers:

// ~/.tooncode/mcp.json
{
  "servers": {
    "my-tool": {
      "command": "python",
      "args": ["path/to/server.py"],
      "env": { "API_KEY": "..." }
    }
  }
}

MCP servers are discovered automatically — their tools become available to the AI alongside built-in tools.

Session & Memory

• Save/Resume — save conversations with /save, resume later with /resume
• Persistent Memory — AI saves important notes across sessions via memory_save
• Edit History — tracks recent file edits, supports /undo

Cross-Platform

| Feature | Windows | Mac | Linux | |---------|:-------:|:---:|:-----:| | All 21 tools | ✅ | ✅ | ✅ | | Shell | PowerShell | bash | bash | | Screenshot | ImageGrab | screencapture | scrot | | Clipboard | Get-Clipboard | pbpaste | xclip | | Browser | ✅ | ✅ | ✅ |

Requirements

• Python 3.10+
• Git
• Node.js 16+ (optional, for npm install method)

Installation

Method 1: One-line install (recommended)

Mac / Linux:

curl -fsSL https://raw.githubusercontent.com/votadev/tooncode/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/votadev/tooncode/main/install.bat -OutFile install.bat; .\install.bat

Method 2: npm

npm install -g @votadev/tooncode

Method 3: Git clone

git clone https://github.com/votadev/tooncode.git ~/.tooncode
pip install -r ~/.tooncode/requirements.txt
python ~/.tooncode/tooncode.py

Add to PATH:

Mac/Linux — add to ~/.bashrc or ~/.zshrc:

alias tooncode="python ~/.tooncode/tooncode.py"

Windows — create tooncode.cmd in a PATH directory:

@echo off
python %USERPROFILE%\.tooncode\tooncode.py %*

Method 4: Just copy the folder

cd /path/to/tooncode
pip install -r requirements.txt
python tooncode.py

Quick Start

tooncode                                    # Start in current directory
tooncode big-pickle                         # Start with specific model
tooncode minimax-m2.5-free ~/my-project     # Model + directory

Configuration

Settings (~/.tooncode/settings.json)

{
  "api_url": "https://opencode.ai/zen/v1/messages",
  "api_key": "public",
  "default_model": "big-pickle",
  "models": [
    { "name": "big-pickle", "context": 200000 },
    { "name": "custom-model", "api_url": "...", "api_key": "..." }
  ],
  "auto_approve": true
}

Custom Skills

Add .md files to ~/.tooncode/skills/<category>/:

---
name: my-skill
description: What this skill does
---

Your prompt template here. Use {{input}} for user input.

Update

cd ~/.tooncode && git pull

License

MIT — Free to use, modify, and distribute.

Made with ❤️ by VotaLab