tui-canvas
MCP server for creating interactive terminal canvases (calendar, document, flight) in AI coding assistants. Supports multiple TUI frameworks and integrates with tmux.
README
TUI Canvas
Framework-agnostic interactive terminal canvases for AI coding assistants.
Works with OpenCode, Claude Code, Cursor, Gemini CLI, Grok CLI, and any MCP-compatible AI harness.
Supports multiple TUI frameworks: OpenTUI (primary), Ink, and more.
Features
- Protocol-first design - Any TUI framework can implement canvases
- Multiple implementations - Same canvas in OpenTUI, Ink, Bubbletea, etc.
- MCP integration - Works with any AI coding assistant via MCP
- tmux integration - Canvases spawn in split panes
Built-in Canvases
- Calendar - Weekly view with events, meeting time picker
- Document - Markdown viewer/editor with text selection
- Flight - Cyberpunk-themed flight booking with seatmaps
Quick Start
Install
# Run without installing globally
bunx tui-canvas list
MCP Configuration
OpenCode (opencode.json in this repo):
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"tui-canvas": {
"type": "local",
"command": ["bun", "run", "packages/mcp/src/index.ts"],
"enabled": true
}
}
}
Other MCP clients:
{
"mcpServers": {
"tui-canvas": {
"command": "bunx",
"args": ["tui-canvas"]
}
}
}
CLI Usage
# List available canvases
bunx tui-canvas list
# Spawn a canvas (must be in tmux)
bunx tui-canvas spawn calendar --scenario display
bunx tui-canvas spawn document --scenario edit --config '{"content": "# Hello"}'
# Use specific implementation
bunx tui-canvas spawn calendar --implementation ink
bunx tui-canvas spawn calendar --implementation opentui
Project Structure
tui-canvas/
├── packages/
│ ├── protocol/ # Protocol specification & types
│ ├── cli/ # Unified CLI
│ ├── mcp/ # MCP server
│ └── runtime/ # Framework adapters
│ ├── opentui/ # OpenTUI runtime (primary)
│ └── ink/ # Ink runtime
│
├── canvases/ # Canvas implementations
│ ├── calendar/
│ │ ├── manifest.json # Canvas manifest
│ │ ├── opentui/ # OpenTUI implementation
│ │ └── ink/ # Ink implementation
│ ├── document/
│ └── flight/
Supported Frameworks
| Framework | Status | Description |
|---|---|---|
| OpenTUI | Primary | TypeScript TUI framework (anomalyco/opentui) |
| Ink | Secondary | React for terminals |
| Bubbletea | Planned | Go TUI framework |
| Textual | Planned | Python TUI framework |
Supported AI Harnesses
| Harness | Detection |
|---|---|
| OpenCode | OPENCODE env |
| Claude Code | CLAUDE_CODE env |
| Cursor | CURSOR_SESSION env |
| Gemini CLI | GEMINI_CLI / GEMINI_API_KEY env |
| Grok CLI | GROK_CLI / XAI_API_KEY env |
| Generic | Fallback for any MCP client |
Creating Canvases
Canvas Manifest
Each canvas has a manifest.json:
{
"id": "my-canvas",
"name": "My Canvas",
"description": "A custom canvas",
"version": "1.0.0",
"scenarios": {
"default": {
"description": "Default scenario",
"configSchema": { ... },
"resultSchema": { ... }
}
},
"implementations": {
"opentui": {
"framework": "opentui",
"reconciler": "solid",
"entrypoint": "./opentui/index.ts"
},
"ink": {
"framework": "ink",
"entrypoint": "./ink/index.tsx"
}
},
"defaultImplementation": "opentui"
}
Protocol
Canvases communicate via Unix sockets using the TUI Canvas Protocol:
Canvas → Controller:
{ type: "ready", scenario: string }
{ type: "selected", data: unknown }
{ type: "cancelled", reason?: string }
Controller → Canvas:
{ type: "update", config: unknown }
{ type: "close" }
Development
# Clone and install
git clone https://github.com/anomalyco/tui-canvas.git
cd tui-canvas
bun install
# Test CLI
bun run packages/cli/src/index.ts list
bun run packages/cli/src/index.ts spawn calendar
# Start MCP server
bun run packages/mcp/src/index.ts
Requirements
License
MIT
Credits
Originally based on Claude Canvas. OpenTUI framework: github.com/anomalyco/opentui
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。