tooner

Tooner is a lightweight MCP wrapper that sits in front of your existing MCP servers. It intercepts and rewrites JSON-heavy outputs into the TOON format, allowing models to understand and process responses with fewer tokens while leaving your server logic untouched. You can point Tooner at any compatible MCP server, and it will transparently proxy traffic, converting large JSON payloads into a more compact representation without changing the underlying data. This makes it particularly useful when you want model-friendly responses without rewriting your MCP server code.

Usage typically involves configuring MCP to launch the Tooner wrapper as a front-end process. You specify the Tooner binary as the command and pass the target MCP server as an argument. For example, if your MCP setup exposes a memory MCP server named memory-mcp-server-go, you can configure Tooner to invoke it, so all traffic flows through Tooner and is reformatted on the way back to the client. You can also use the NPX workflow to run Tooner directly without a local install, and you can configure specific environment variables for your backend (such as database connections or logging) when needed.

Prerequisites:

• A supported operating system (macOS, Linux, Windows) and a Go toolchain if you plan to build from source.
• Optional: Homebrew for macOS/Linux, Node.js if you prefer NPX workflows, or access to release binaries.

Install options (choose one):

1. Homebrew (macOS/Linux):

# Install
brew install chaindead/tap/tooner

# Update
brew upgrade chaindead/tap/tooner

2. NPX (no local install, cross-platform):

npx -y @chaindead/tooner

3. From Releases (binary install):

• MacOS/Linux (binary in releases)

# Download the appropriate binary, then place in your PATH

4. From Source (go toolchain required):

# Ensure Go is installed and in PATH
GO111MODULE=on

# Install the latest Tooner binary from source
go install github.com/chaindead/tooner@latest

After installing, you can run Tooner and configure MCP to use it as a front-end wrapper (see the MCP configuration examples in the README).

Tips and notes:

• Tooner proxies traffic and converts JSON outputs to TOON format. If your MCP server emits complex or large JSON payloads, you’ll see reduced token usage in model interactions.
• Example configuration shows how to place Tooner in front of an existing MCP server (e.g., memory-mcp-server-go) by setting the command to tooner and passing the target as an argument.
• For more complex backends (like Postgres-backed MCP), you can run Tooner with additional environment variables such as DATABASE_URI or TOONER_LOG_PATH to control connections and logging.
• If you encounter path or permission issues after installing from releases, ensure the binary is executable (chmod +x) and that your PATH includes the directory containing the binary.
• NPX usage is handy for quick testing or ephemeral setups but may require network access to fetch the package on first run.