Follow these steps to run the MCP agent and toolkit in a temporary Meshagent Room for experimentation or development. All resources will be automatically removed when the room closes.
  1. Install Meshagent 
    pip install "meshagent[all]"
  2. Sign Up & Authenticate
    Create an account and configure authentication, following the Meshagent CLI Getting Started Guide.
  3. Start a Test MCP Server Tool in a Room 
    meshagent service test --room=test --role=agent --image=meshagent/mcp-node-code-sandbox:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-node-code-sandbox-service-test
    • This command launches a Meshagent Room (test) with a disposable MCP server.
    • If the Room goes inactive, all resources (including the MCP server) will be automatically cleaned up.
  4. Start a Chatbot in the Room with the Node Code Sandbox Toolkit 
    meshagent chatbot join --room=test --agent-name=mcp-node-code-sandbox --name=mcp-node-code-sandbox --toolkit=mcp-node-code-sandbox
    • This launches a chatbot agent in the same Room, enabled with the mcp-node-code-sandbox toolkit.
    • You can combine multiple toolkits in the same agent and room.
    • A link to the room will be printed in the command output.
  5. Access the Room and Interact
    • Open the Room link provided by the above command in a web browser.
    • Send a message to the agent to interact with the MCP server tools (such as running code, managing sandboxes, etc).
All cloud resources are tied to the Room lifecycle—no cleanup needed!

​
Project Level Deployment

For persistent, production-grade deployments, register the MCP Server and chatbot at the project level—so they auto-join every time a new Room is created. This eliminates the need to manually start services for each session, or to run the server/chatbot locally.
  1. Create a Persistent MCP Service for Your Project 
    meshagent service create --role=agent --image=meshagent/mcp-node-code-sandbox:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-node-code-sandbox-service
  2. Create a Persistent Chatbot Service with the Code Sandbox Toolkit 
    meshagent service create --image="meshagent/cli:latest" --port="num=9001 path=/agent liveness=/ type=meshagent.callable participant_name=mcp-node-code-sandbox-chatbot" --env="MESHAGENT_PORT=9001" --name="mcp-node-code-sandbox-chatbot-service" --command="meshagent chatbot service --agent-name=mcp-node-code-sandbox-chatbot --toolkit=mcp-node-code-sandbox"
Once created, these services automatically join all new rooms in the project—providing consistent tools and chatbot capability with zero local setup.

​
Tools Available

The following tools are available via the Node.js Sandbox MCP Server, enabled in Meshagent Rooms via this integration:
ToolShort Description
get_dependency_typesGiven an array of npm package names (and optional versions), fetch whether each package ships its own TypeScript definitions or has a corresponding @types/… package, and return the raw .d.ts text.
run_jsInstall npm dependencies and run JavaScript code inside a running sandbox container.
run_js_ephemeralRun a JavaScript snippet in a temporary disposable container with optional npm dependencies, then automatically clean up.
sandbox_execExecute one or more shell commands inside a running sandbox container.
sandbox_initializeStart a new isolated Docker container running Node.js.
sandbox_stopTerminate and remove a running sandbox container.
search_npm_packagesSearch for npm packages by a search term and get their name, description, and a README snippet.

​
Tools Details

​
get_dependency_types

Given an array of npm package names (and optional versions), fetch whether each package ships its own TypeScript definitions or has a corresponding @types/… package, and return the raw .d.ts text. Useful when you’re about to run a Node.js script against an unfamiliar dependency and want to inspect what APIs and types it exposes.
ParametersTypeDescription
dependenciesarray

​
run_js

Install npm dependencies and run JavaScript code inside a running sandbox container.
  • After running, you must manually stop the sandbox to free resources.
  • The code must be valid ESModules (import/export syntax).
  • Best for complex workflows where you want to reuse the environment across multiple executions.
  • Read/write via the ./files directory for persistence.
ParametersTypeDescription
codestringJavaScript code to run inside the container.
container_idstringDocker container identifier
dependenciesarray (optional)npm dependencies to install before running. Each must have name and version.
listenOnPortnumber (optional)Expose this port to the host if set.

​
run_js_ephemeral

Run a JavaScript snippet in a temporary disposable container with optional npm dependencies, and auto-clean up on completion.
  • Code must be valid ESModules.
  • Ideal for simple one-shot runs.
  • All IO (including images and binaries) must use the ./files directory.
Example: 
import fs from "fs/promises";
await fs.writeFile("./files/hello.txt", "Hello world!");
console.log("Saved ./files/hello.txt");
ParametersTypeDescription
codestringJavaScript code to run.
dependenciesarray (optional)npm dependencies to install.
imagestring (optional)Docker image to use, e.g., node:lts-slim, mcr.microsoft.com/playwright:v1.52.0-noble, alfonsograziano/node-chartjs-canvas:latest.

​
sandbox_exec

Execute one or more shell commands inside a running sandbox container. (Requires an initialized sandbox.)
ParametersTypeDescription
commandsarrayShell commands to run.
container_idstring

​
sandbox_initialize

Start a new isolated Docker container running Node.js.
ParametersTypeDescription
imagestring (optional)
portnumber (optional)Map this port.

​
sandbox_stop

Terminate and remove a running sandbox container.
ParametersTypeDescription
container_idstring

​
search_npm_packages

Search for npm packages by a search term—get name, description, and README snippet.
ParametersTypeDescription
searchTermstringSearch term (may include + to combine concepts).
qualifiersobject (optional)Qualifiers to filter results (by author, scope, etc).