This MCP server provides the BMAD Agent FastMCP Service, exposing a modular set of agents, workflows, and LLM modes that integrate with Cursor IDE. It enables you to run a local BMAD MCP instance, switch between built-in and external LLMs, and orchestrate tasks through a suite of prebuilt tools and workflows.

How to use

You run the BMAD Agent MCP server locally and connect your MCP client to it. Start the service with Python, then interact through the client to list agents, start workflows, or switch LLM modes. Use the provided tools and workflows to automate common tasks across analysis, architecture, development, product management, and QA.

Key usage patterns you can perform include listing all professional agents, activating a chosen agent, invoking an agent to complete a task, listing and triggering workflows, and switching between Cursor’s built-in LLM and the external DeepSeek API for more advanced reasoning.

Example interactions you might perform inside your MCP client include: list_agents to view available agents, call_agent_with_llm with a specific task, and switch_llm_mode to switch the LLM backend. The system also exposes status queries to monitor health and progress.

How to install

Prerequisites: you need Python installed and access to install Python packages.

pip install -r requirements.txt

Create a local environment file by copying the template and optionally adjust settings.

cp .env.example .env

Optional edits in .env

USE_BUILTIN_LLM=true

DEEPSEEK_API_KEY=your_key

Start the MCP server using the main Python script.

python bmad_agent_mcp.py

## Configuration and integration

Integrate with Cursor IDE by configuring the MCP server in your Cursor settings. The server runs locally and uses Python as the runtime for the main script.

{ "mcpServers": { "bmad_agent": { "command": "python", "args": ["bmad_agent_mcp.py"], "cwd": ".", "env": { "PYTHONPATH": ".", "USE_BUILTIN_LLM": "true", "PYTHONIOENCODING": "utf-8", "PYTHONUNBUFFERED": "1" } } } }

## Environment and security notes

Environment variables control LLM usage and encoding. Provide your own DeepSeek API key if you plan to use the external LLM. Keep API keys and secrets secure, and avoid committing .env files to version control.

## Examples of available tools and actions

The server exposes a collection of tools to manage agents, workflows, tasks, and templates. You can discover and use these through your MCP client.

## Available tools

### list\_agents

Return a list of all professional agents available in the system.

### get\_agent\_details

Fetch detailed information for a specific agent by its ID.

### activate\_agent

Activate or enable a specified agent to start handling tasks.

### call\_agent\_with\_llm

Invoke an agent to perform a task using LLM-based reasoning.

### list\_workflows

List all defined workflows in the MCP environment.

### start\_workflow

Start a selected workflow by its ID.

### get\_workflow\_status

Check the current status of an active workflow.

### advance\_workflow\_step

Advance the workflow to the next step.

### switch\_llm\_mode

Switch between built-in Cursor LLM and external DeepSeek API.

### get\_llm\_mode\_info

Retrieve current LLM mode information.

### get\_system\_status

Query the overall health and status of the MCP server.

### list\_tasks

List all available tasks in the MCP system.

### execute\_task

Execute a specific task by its ID.

### list\_templates

List all available templates for reuse in tasks or workflows.

### get\_template

Retrieve the content of a specific template by name.