aseprite-mcp
MCP server for Aseprite — create, edit, and export pixel art sprites, animations, and sprite sheets from any AI assistant.
README
aseprite-mcp
An MCP server for Aseprite — create, edit, and export pixel art sprites, animations, and sprite sheets from any AI assistant.
<p align="center"> <img src="assets/demo.gif" alt="Demo — pixel art knight walking animation created with aseprite-mcp" width="256" /> </p>
Drawn and animated entirely via aseprite-mcp tools — no manual pixel editing!
Features
- 43 tools across 11 categories
- Native drawing with configurable brush thickness via
app.useTool() - Pixel-perfect algorithms (Bresenham line, midpoint circle) for thin strokes
- Cross-platform — Windows, macOS, Linux
- Secure — Lua injection prevention, sandboxed execution
- Zero runtime dependencies beyond
@modelcontextprotocol/sdk
Tool Overview
| Category | Tools | Examples |
|---|---|---|
| Sprites | 6 | create_sprite, resize_sprite, crop_sprite |
| Layers | 4 | add_layer, set_layer_properties, list_layers |
| Frames | 4 | add_frame, set_frame_duration, list_frames |
| Tags | 3 | create_tag, remove_tag, list_tags |
| Drawing | 7 | draw_rect, draw_circle, draw_line, draw_ellipse, fill_area, outline, draw_pixels |
| Transform | 5 | replace_color, flip_sprite, rotate_sprite, flatten_layers, merge_down |
| Palette | 4 | get_palette, set_palette_colors, load_palette, resize_palette |
| Cels | 3 | move_cel, set_cel_opacity, clear_cel |
| Export | 3 | export_sprite_sheet, export_frame, export_layers |
| Slices | 2 | create_slice, remove_slice |
| Utility | 2 | run_script, get_aseprite_version |
📖 Full parameter reference: docs/API.md
Quick Start
Requirements
- Node.js ≥ 18
- Aseprite ≥ 1.3 (aseprite.org)
Install
# npm global
npm install -g aseprite-mcp
# or from source
git clone https://github.com/ayigityol/aseprite-mcp.git
cd aseprite-mcp && npm install && npm run build
Configure
<details> <summary><strong>GitHub Copilot CLI</strong> (~/.copilot/mcp-config.json)</summary>
{
"mcpServers": {
"aseprite": {
"type": "local",
"command": "node",
"tools": ["*"],
"args": ["/path/to/aseprite-mcp/build/index.js"],
"env": { "ASEPRITE_PATH": "/path/to/aseprite" }
}
}
}
</details>
<details> <summary><strong>Claude Desktop</strong> (claude_desktop_config.json)</summary>
{
"mcpServers": {
"aseprite": {
"command": "node",
"args": ["/path/to/aseprite-mcp/build/index.js"],
"env": { "ASEPRITE_PATH": "/path/to/aseprite" }
}
}
}
</details>
<details> <summary><strong>VS Code / Cursor</strong> (.vscode/mcp.json)</summary>
{
"servers": {
"aseprite": {
"command": "node",
"args": ["/path/to/aseprite-mcp/build/index.js"],
"env": { "ASEPRITE_PATH": "/path/to/aseprite" }
}
}
}
</details>
<details> <summary><strong>Installed globally via npm</strong></summary>
{
"mcpServers": {
"aseprite": {
"command": "aseprite-mcp"
}
}
}
</details>
Environment Variables
| Variable | Description |
|---|---|
ASEPRITE_PATH |
Path to Aseprite executable. Auto-detected if not set. |
DEBUG |
Set "true" for verbose stderr logging. |
Auto-detection searches standard install paths on all platforms, plus system PATH.
Architecture
sequenceDiagram
participant Client as MCP Client
participant Server as aseprite-mcp
participant Aseprite as Aseprite CLI
Client->>Server: Tool call (JSON-RPC via stdio)
Server->>Server: Generate Lua script
Server->>Aseprite: aseprite -b --script temp.lua
Aseprite->>Aseprite: Execute Lua (headless)
Aseprite-->>Server: stdout: __RESULT__{"success":true, ...}
Server->>Server: Parse JSON, cleanup temp file
Server-->>Client: Tool result
graph LR
A[Tool Call] --> B{Thickness > 1?}
B -->|Yes| C[app.useTool<br/>Native Brush]
B -->|No| D[image:drawPixel<br/>Pixel Algorithms]
C --> E[Save & Return]
D --> E
All operations run headless — no GUI window is opened.
Docker
# Build
docker build -t aseprite-mcp .
# Run (mount your Aseprite binary + working directory)
docker run --rm -i \
-v /path/to/aseprite:/usr/local/bin/aseprite:ro \
-v ./sprites:/sprites \
aseprite-mcp
# Or use Docker Compose
docker compose up
Aseprite must be mounted into the container. The image packages only the MCP server.
Development
npm run build # Compile TypeScript
npm run watch # Recompile on changes
npm test # Run 74 unit tests (vitest)
npm run test:watch # Watch mode
npm run inspector # MCP Inspector for interactive testing
Security
luaEscape()— prevents Lua injection by escaping\,",\n,\rluaPath()— normalizes and escapes file pathsexecFile()— argument arrays, no shell interpolationpcall()wrapping — all generated Lua scripts have error handlers- Local only — no data sent to external services
Troubleshooting
| Problem | Solution |
|---|---|
| "Aseprite executable not found" | Set ASEPRITE_PATH env var to the full path |
| Tool returns Aseprite error | Set DEBUG=true to see stderr output |
| "Layer not found" / "Tag not found" | Names are case-sensitive — use list_layers or list_tags first |
| Sprite won't save | Ensure the output directory exists |
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。