TOONer
MCP proxy that wraps any MCP server and transparently converts JSON responses in tools/call to TOON format — a token-efficient alternative to JSON optimized for LLMs (~40% fewer tokens).
README
Tooner
Tooner is a lightweight MCP wrapper that keeps your server setup unchanged, but makes model-facing responses cleaner and shorter. It rewrites JSON-heavy outputs into TOON format so models spend fewer tokens on syntax noise (~40% fewer tokens).
What it does
- Works with your existing MCP servers (no server rewrite required)
- Keeps the same data, but in a format models read more efficiently
- Reduces token usage on large JSON responses
- Handles many messy JSON-like outputs and still returns useful model-ready content
Installation
Homebrew
You can install a binary release on macOS/Linux using brew:
# Install
brew install chaindead/tap/tooner
# Update
brew upgrade chaindead/tap/tooner
NPX
You can run the latest version directly using npx (supports macOS, Linux, and Windows):
npx -y @chaindead/tooner
When using NPX Configure MCP (Cursor example) becomes:
{
"mcpServers": {
"telegram": {
"command": "npx",
"args": ["-y", "@chaindead/tooner", "memory-mcp-server-go"]
}
}
}
From Releases
MacOS
<details>
Note: The commands below install to
/usr/local/bin. To install elsewhere, replace/usr/local/binwith your preferred directory in your PATH.
First, download the archive for your architecture:
# For Intel Mac (x86_64)
curl -L -o tooner.tar.gz https://github.com/chaindead/tooner/releases/latest/download/tooner_Darwin_x86_64.tar.gz
# For Apple Silicon (M1/M2)
curl -L -o tooner.tar.gz https://github.com/chaindead/tooner/releases/latest/download/tooner_Darwin_arm64.tar.gz
Then install the binary:
# Extract the binary
sudo tar xzf tooner.tar.gz -C /usr/local/bin
# Make it executable
sudo chmod +x /usr/local/bin/tooner
# Clean up
rm tooner.tar.gz
</details>
Linux
<details>
Note: The commands below install to
/usr/local/bin. To install elsewhere, replace/usr/local/binwith your preferred directory in your PATH.
First, download the archive for your architecture:
# For x86_64 (64-bit)
curl -L -o tooner.tar.gz https://github.com/chaindead/tooner/releases/latest/download/tooner_Linux_x86_64.tar.gz
# For ARM64
curl -L -o tooner.tar.gz https://github.com/chaindead/tooner/releases/latest/download/tooner_Linux_arm64.tar.gz
Then install the binary:
# Extract the binary
sudo tar xzf tooner.tar.gz -C /usr/local/bin
# Make it executable
sudo chmod +x /usr/local/bin/tooner
# Clean up
rm tooner.tar.gz
</details>
Windows
<details>
Windows
- Download the latest release for your architecture:
- Extract the
.zipfile - Add the extracted directory to your PATH or move
tooner.exeto a directory in your PATH </details>
From Source
Requirements:
- Go 1.26 or later
- GOBIN in PATH
go install github.com/chaindead/tooner@latest
Configure MCP
Setup is the only thing you need to do: in MCP config, shift the command by one argument and insert tooner at the beginning.
Before:
{
"mcpServers": {
"memory": {
"command": "memory-mcp-server-go"
}
}
}
After:
{
"mcpServers": {
"memory": {
"command": "tooner",
"args": ["memory-mcp-server-go"]
}
}
}
Another example - Postgres MCP:
{
"mcpServers": {
"postgres": {
"command": "tooner",
"args": [
"uvx",
"postgres-mcp",
"--access-mode=unrestricted"
],
"env": {
"DATABASE_URI": "...",
"TOONER_LOG_PATH": "/tmp/tooner.log"
}
}
}
}
NPX setup:
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["-y", "@chaindead/tooner", "memory-mcp-server-go"]
}
}
}
That's it. From there, tooner launches your MCP server as a subprocess and transparently proxies traffic, replacing JSON with TOON.
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。