The following guide walks you through setting up, testing, and cleaning up an MCP agent and chatbot inside a Meshagent Room.
  1. Install Meshagent and All Extras 
    pip install "meshagent[all]"
  2. Sign Up and Authenticate
    • Visit: Getting Started
    • Follow instructions to sign in and set up your auth.
  3. Start MCP Server as a Room Agent for Testing 
    meshagent service test --room=test --role=agent --image=meshagent/mcp-everything:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-everything-service-test
    • This command launches your test MCP server inside a Meshagent Room (test). The service is available only while the Room exists.
  4. Start a Chatbot in the Room Using the Everything Toolkit 
    meshagent chatbot join --room=test --agent-name=mcp-everything --name=mcp-everything --toolkit=mcp-everything
    • You may use multiple toolkits for the same agent in a Room.
    • The output gives you a URL to the Room.
  5. Try It Out
    • Open the Room link from the step above in your browser.
    • Send a message to the agent—directly interact with the MCP Server tools via the chatbot.
Note: When the Room closes (e.g., by going inactive), the MCP server and chatbot are automatically removed.

​
Project Level Deployment (Persistent Services)

To have the MCP server tools and chatbot available in every Room for a given Meshagent project (ideal for production):
  1. Create a Persistent MCP Server Agent Service 
    meshagent service create --role=agent --image=meshagent/mcp-everything:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-everything-service
  2. Create a Persistent Chatbot Service in the Project 
    meshagent service create --image="meshagent/cli:latest" --port="num=9001 path=/agent liveness=/ type=meshagent.callable participant_name=mcp-everything-chatbot" --env="MESHAGENT_PORT=9001" --name="mcp-everything-chatbot-service" --command="meshagent chatbot service --agent-name=mcp-everything-chatbot --toolkit=mcp-everything"
With these persistent services, the MCP agent and the chatbot will automatically join every new Room created in your project, eliminating the need to run them locally or launch them each time.

​
Tools Available

​
Tools Provided by This Server

Tool NameShort Description
addAdds two numbers
annotatedMessageDemonstrates how annotations can be used for metadata
echoEchoes back the input
getResourceReferenceReturns a resource reference usable by MCP clients
getTinyImageReturns the MCP_TINY_IMAGE
longRunningOperationDemo: long running operation with progress updates
printEnvPrints environment variables for debugging
sampleLLMSamples from an LLM via MCP’s sampling feature

​
Tools Details

​
add

Adds two numbers
  • Parameters:
    • a (number): First number
    • b (number): Second number

​
annotatedMessage

Provides example annotations (metadata) for content
  • Parameters:
    • messageType (string): Type for annotation
    • includeImage (boolean, optional): Attach an example image

​
echo

Echoes back your input
  • Parameters:
    • message (string): Message to echo

​
getResourceReference

Returns a resource reference for MCP clients
  • Parameters:
    • resourceId (number): Resource ID (1-100)

​
getTinyImage

Returns the MCP_TINY_IMAGE
  • Parameters: None

​
longRunningOperation

Demonstrates long-running operations and progress updates
  • Parameters:
    • duration (number, optional): Duration in seconds
    • steps (number, optional): Number of steps

​
printEnv

Prints all environment variables
  • Parameters: None

​
sampleLLM

Sample from an LLM using MCP
  • Parameters:
    • prompt (string): Prompt for the LLM
    • maxTokens (number, optional): Max tokens to generate

​
Meshagent & Resources

For more details on MCP and Model Context Protocol, see the Anthropic MCP announcement.
Start building collaborative, tool-driven agents today using Meshagent and the Everything MCP Server!