Follow these steps to spin up a temporary (ephemeral) MCP server and connect a chatbot in a Meshagent Room for testing and development:

​
Prerequisites

​
Required Environment Variables

  • DOIT_API_KEY: your DoiT API key

​
Steps

  1. Install Meshagent CLI + All Dependencies 
    pip install "meshagent[all]"
  2. Authenticate with Meshagent 
    meshagent login
    (Follow prompts to authenticate.)
  3. Launch MCP Server as an Agent in a Room 
    meshagent service test --room=test --role=agent --image=meshagent/mcp-doit:latest --env DOIT_API_KEY=your_doit_api_key --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-doit-service-test
    • Starts a Meshagent Room (“test”) with the MCP server using your API key.
    • The server process and tools will be removed when the room goes inactive.
  4. Start Chatbot in the Same Room with DoiT Toolkit 
    meshagent chatbot join --room=test --agent-name=mcp-doit --name=mcp-doit --toolkit=mcp-doit
    • Joins a chatbot (mcp-doit) connected to the MCP tools.
    • Note: Output will provide a direct link to the Room.
  5. Try it Out
    • Visit the Room link in your web browser.
    • Send chat commands to the agent to interact with the MCP Server tools.

​
Project Level Deployment: Persistent MCP Server in All Project Rooms

For production/always-on deployments, create persistent MCP services tied to your Meshagent project. These services will auto-join any new Room, making MCP tools and chatbot always available without extra setup.

​
Steps

  1. Create Persistent MCP Server Service 
    meshagent service create --role=agent --image=meshagent/mcp-doit:latest --env DOIT_API_KEY=your_doit_api_key --env MESHAGENT_PORT=8001 --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-doit-service
  2. Create Persistent Chatbot Service with Toolkit 
    meshagent service create --image="meshagent/cli:latest" --port="num=9001 path=/agent liveness=/ type=meshagent.callable participant_name=mcp-doit-chatbot" --env="MESHAGENT_PORT=9001" --name="mcp-doit-chatbot-service" --command="meshagent chatbot service --agent-name=mcp-doit-chatbot --toolkit=mcp-doit"
Rooms in this project will automatically get the MCP server tools and chatbot every time a Room is created.

​
Tools Available

The DoiT MCP Server exposes the following tools to chatbots and Meshagent:
ToolShort Description
get_anomaliesList anomalies detected in cloud costs
get_anomalyGet a specific anomaly by ID
get_cloud_incidentGet a specific cloud incident by ID
get_cloud_incidentsGet cloud incidents
get_dimensionGet a specific Cloud Analytics dimension by type and ID
get_report_resultsGet the results of a specific report by ID
list_dimensionsLists Cloud Analytics dimensions that your account has access to.
list_reportsLists Cloud Analytics reports that your account has access to
run_queryRuns a report query with the specified configuration without persisting it.
validate_userValidates the current API user and returns domain and email information

​
Tool Details

​
get_anomalies

List anomalies detected in cloud costs
ParameterTypeDescription
filterstring optionalFilter string in format ‘key:valuekey:value’. Multiple values for same key are treated as OR, different keys as AND.
pageTokenstring optionalToken for pagination. Use this to get the next page of results.

​
get_anomaly

Get a specific anomaly by ID
ParameterTypeDescription
idstringanomaly ID

​
get_cloud_incident

Get a specific cloud incident by ID
ParameterTypeDescription
idstringincident ID

​
get_cloud_incidents

Get cloud incidents
ParameterTypeDescription
filterstring optionalFilter string in format ‘key:valuekey:value’. Multiple values for same key are treated as OR, different keys as AND. Example: ‘platform:google-cloudstatus:active’ or ‘platform:google-cloudplatform:amazon-web-services’
pageTokenstring optionalToken for pagination. Use this to get the next page of results.
platformstring optionalplatform name

​
get_dimension

Get a specific Cloud Analytics dimension by type and ID
ParameterTypeDescription
idstringDimension id
typestringDimension type

​
get_report_results

Get the results of a specific report by ID
ParameterTypeDescription
idstringThe ID of the report to retrieve results for

​
list_dimensions

Lists Cloud Analytics dimensions that your account has access to. Use this tool to get the dimensions that you can use in the run_query tool. Use filter to narrow down the results.
ParameterTypeDescription
filterstring optionalFilter string (optional) in format ‘key:valuekey:value’. Multiple values for same key are treated as OR, different keys as AND. The fields eligible for filtering are: type, label, key. Use the filter parameter only if you know the exact value of the key, otherwise the filter should be empty.
pageTokenstring optionalToken for pagination. Use this to get the next page of results.

​
list_reports

Lists Cloud Analytics reports that your account has access to
ParameterTypeDescription
filterstring optionalFilter string in format ‘key:valuekey:value’. Multiple values for same key are treated as OR, different keys as AND. Possible filter keys: reportName, owner, type, updateTime, use the filter property only if you know for sure the value is a valid filter key, do not guess it.
pageTokenstring optionalToken for pagination. Use this to get the next page of results.

​
run_query

Runs a report query with the specified configuration without persisting it.
Fields that are not populated will use their default values if needed.
Use the dimension tool before running the query to get the list of dimensions and their types.
If possible, use timeRange instead of customTimeRange when no specific dates are given. Example for cost report: 
{
  "config": {
    "dataSource": "billing",
    "metric": {"type": "basic", "value": "cost"},
    "timeRange": {"mode": "last", "amount": 1, "unit": "month", "includeCurrent": true},
    "group": [
      {
        "id": "service_description",
        "type": "fixed",
        "limit": {"metric": {"type": "basic", "value": "cost"}, "sort": "desc", "value": 10}
      }
    ]
  }
}
ParameterTypeDescription
configobjectThe configuration for the query, including dimensions, metrics, filters, etc.

​
validate_user

Validates the current API user and returns domain and email information

​
Additional Resources