The following steps walk you through launching an MCP server and chatbot agent within a Meshagent Room for cloud-based testing. Both the tool and agent automatically stop when the room closes.
  1. Install Meshagent & Full Feature Set
    pip install "meshagent[all]"
    
  2. Sign Up & Authenticate
  3. Start a Test MCP Server as a Cloud Room Service
    meshagent service test --room=test --role=agent --image=meshagent/mcp-memory:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-memory-service-test
    
    • This will launch a Meshagent Room in the cloud named test with an MCP server running as a service in that room.
    • The MCP server (mcp-memory-service-test) and its tools will be live only as long as the room is active.
  4. Add a Chatbot Agent with the MCP Memory Toolkit
    meshagent chatbot join --room=test --agent-name=mcp-memory --name=mcp-memory --toolkit=mcp-memory
    
    • This launches a chatbot linked to the MCP memory toolkit within the same room.
    • Multiple toolkits may be assigned to a single agent.
    • The command output provides a browser link to join the room interactively.
  5. Test the Tools
    • Visit the room link in your browser.
    • Send a message to the agent to interact with the MCP server tools directly in the room chat interface.

Project Level Deployment

For production and persistent deployments: Register services at the Meshagent project level, so every newly created Room automatically gains access to the MCP server tools and a chatbot—no need to run them locally or as separate processes. This makes MCP toolkits and agents available across all rooms.
  1. Create the MCP Memory Service Project-Wide
    meshagent service create --role=agent --image=meshagent/mcp-memory:latest --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-memory-service
    
  2. Create the MCP Memory Chatbot Project-Wide
    meshagent service create --image="meshagent/cli:latest" --port="num=9001 path=/agent liveness=/ type=meshagent.callable participant_name=mcp-memory-chatbot" --env="MESHAGENT_PORT=9001" --name="mcp-memory-chatbot-service" --command="meshagent chatbot service --agent-name=mcp-memory-chatbot --toolkit=mcp-memory"
    

Tools Available

Tools provided by this MCP ServerShort Description
add_observationsAdd new observations to existing entities in the knowledge graph
create_entitiesCreate multiple new entities in the knowledge graph
create_relationsCreate multiple new relations between entities in the knowledge graph.
delete_entitiesDelete multiple entities and their associated relations from the knowledge graph
delete_observationsDelete specific observations from entities in the knowledge graph
delete_relationsDelete multiple relations from the knowledge graph
open_nodesOpen specific nodes in the knowledge graph by their names
read_graphRead the entire knowledge graph
search_nodesSearch for nodes in the knowledge graph based on a query

Tools Details

Tool: add_observations

Add new observations to existing entities in the knowledge graph
ParametersTypeDescription
observationsarray

Tool: create_entities

Create multiple new entities in the knowledge graph
ParametersTypeDescription
entitiesarray

Tool: create_relations

Create multiple new relations between entities in the knowledge graph. Relations should be in active voice
ParametersTypeDescription
relationsarray

Tool: delete_entities

Delete multiple entities and their associated relations from the knowledge graph
ParametersTypeDescription
entityNamesarrayAn array of entity names to delete

Tool: delete_observations

Delete specific observations from entities in the knowledge graph
ParametersTypeDescription
deletionsarray

Tool: delete_relations

Delete multiple relations from the knowledge graph
ParametersTypeDescription
relationsarrayAn array of relations to delete

Tool: open_nodes

Open specific nodes in the knowledge graph by their names
ParametersTypeDescription
namesarrayAn array of entity names to retrieve

Tool: read_graph

Read the entire knowledge graph

Tool: search_nodes

Search for nodes in the knowledge graph based on a query
ParametersTypeDescription
querystringThe search query to match against entity names, types, and observation content