wassette

Wassette is a security-focused runtime that enables MCP clients to load and execute WebAssembly Components inside a Wasmtime sandbox. It acts as an intermediary that allows agents to discover, load, and interact with Wasm Components securely, extending AI agents with new tools without leaving the chat interface. Once Wassette is running, you can register it with your MCP client (for example, via a quick-start guide provided by the Wassette project) and begin loading components from OCI registries or other component sources. After a component is loaded, you can query it or have your agent request tasks that the component can perform, such as time-related queries, data processing, or tool integrations, all executed in isolation within the sandbox.

Typical usage flow: 1) Install Wassette and start the service; 2) Register Wassette with your MCP client using the provided quick-start or client guides; 3) Instruct your agent to load a Wasm Component by referencing its source (for example an OCI image); 4) Interact with the component via natural language prompts and the MCP client, enabling the agent to call tool functions exposed by the component. The example in the project shows loading a time component and querying the current time, demonstrating how a component can be used to extend agent capabilities securely.

Prerequisites:

• Docker or a compatible container runtime installed on your system
• Basic command-line tooling (bash, curl)

Option A: Run Wassette via Docker (recommended for quick testing)

1. Ensure Docker is running on your machine.
2. Start Wassette from the image (adjust image name if needed):

docker run -i microsoft/wassette

3. Follow the on-screen prompts or reference the project documentation for configuring MCP clients and the Wassette server address.

Option B: Install via the official quick-start script (Linux/macOS)

1. Install prerequisites: curl, bash
2. Run the quick-start installer:

curl -fsSL https://raw.githubusercontent.com/microsoft/wassette/main/install.sh | bash

3. After installation, start Wassette according to the installation guide and configure MCP clients as documented in the Quick Start guide.

Option C: Complete installation guide (all platforms)

1. Visit the Installation Guide for detailed steps across Windows, macOS, Linux, Homebrew, Nix, and Docker environments: https://microsoft.github.io/wassette/latest/installation.html
2. Follow platform-specific steps to install binaries or set up container-based execution, then start the Wassette service and register it with your MCP clients.

Tips and known considerations:

• Wassette runs components in a Wasmtime sandbox for strong isolation; ensure your MCP client is configured to communicate with the Wassette runtime as per the Quick Start guide.
• If you are using the Docker route, you may need to map ports or volumes depending on your MCP client integration needs.
• When loading components from OCI registries, ensure proper authentication is configured if the registry requires it.
• Check for updates regularly; the project notes emphasize ongoing development, so environments and commands may evolve.
• Environment variables you might encounter include registry authentication tokens, component source URLs, and MCP client endpoints; refer to the installation guide for a full list and example values.