home / mcp / agent studio mcp server

Agent Studio MCP Server

Exposes Agent Studio workflows as MCP-callable tools for listing, inspecting, creating, and managing workflows and agents.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "jasonmeverett-cloudera-ai-agent-studio-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/cloudera-ai-agent-studio-mcp",
        "run",
        "serve.py"
      ],
      "env": {
        "CDSW_APIV2_KEY": "YOUR-TOKEN-HERE",
        "AGENT_STUDIO_DOMAIN": "https://your-studio.CDSW_DOMAIN.cldr.work"
      }
    }
  }
}

Agent Studio MCP Server exposes your Agent Studio instance as a set of callable tools so MCP-aware clients can list, inspect, and orchestrate workflows on the fly. It acts as a lightweight bridge that lets you manage workflows, agents, and tasks from a centralized MCP client like Claude Desktop.

How to use

You interact with the Agent Studio MCP Server through an MCP client. Use the available tools to list workflows, fetch details, create new workflows, convert existing workflows to conversational mode, and add managers or agents to workflows. This lets you build, inspect, and modify workflows from a single interface without directly touching Agent Studio.

How to install

Prerequisites: ensure you have the MCP runtime available on your machine and that you can run a local stdio MCP server. You will configure a local MCP entry in your client to point at the Agent Studio runtime.

Step 1. Create a directory for the Agent Studio MCP server files.

Step 2. Place the Agent Studio MCP runtime scripts in that directory so you can reference serve.py from the working directory. If you already have the files, skip this step.

Step 3. In your client configuration, define the MCP server as shown below. This enables the client to start the bridge locally and communicate with Agent Studio.

Step 4. Start the MCP server via the runtime command shown in the example configuration.

Additional content

Configuration and environment variables: You must provide environment variables that authenticate and configure your Agent Studio domain. The example below shows two required environment variables: CDSW_APIV2_KEY for API authentication and AGENT_STUDIO_DOMAIN to specify your Agent Studio domain.

{
  "mcpServers": {
    "agent-studio": {
      "type": "stdio",
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/cloudera-ai-agent-studio-mcp",
        "run",
        "serve.py"
      ],
      "env": {
        "CDSW_APIV2_KEY": "YOUR-TOKEN-HERE",
        "AGENT_STUDIO_DOMAIN": "https://your-studio.CDSW_DOMAIN.cldr.work"
      }
    }
  }
}

Available tools

list_current_workflows

Returns every workflow’s ID and name to populate a UI dropdown of existing projects.

get_workflow_information

Fetches full JSON metadata for a single workflow to inspect tasks, agents, and process settings.

create_workflow

Creates a brand-new blank workflow with a given name and description to automate project scaffolding.

make_workflow_conversational

Adds a Conversational Task to a workflow and marks it as conversational, enabling a chat-first experience.

add_manager_agent_to_workflow

Creates a manager agent and wires it as the workflow’s supervisor for hierarchical workflows.

add_agent_to_workflow

Adds one or more agents to a workflow to expand the team with domain specialists.