Build a Voice-Controlled GitHub Agent in Python (MCP + Vision Agents)

uv  init  github-voice-agent

cd  github-voice-agent

# Install Vision Agents and the OpenAI Python plugin

uv  add  "vision-agents[getstream, openai]"

import  logging

import  os



from  dotenv  import  load_dotenv



from  vision_agents.core.agents  import  Agent,  AgentLauncher

from  vision_agents.core  import  cli

from  vision_agents.core.mcp  import  MCPServerRemote

from  vision_agents.plugins.openai.openai_realtime  import  Realtime

from  vision_agents.plugins  import  getstream

from  vision_agents.core.events  import  CallSessionParticipantJoinedEvent

from  vision_agents.core.edge.types  import  User



# Load environment variables from .env file

load_dotenv()



# Set up logging

logging.basicConfig(level=logging.INFO)

logger  =  logging.getLogger(__name__)



async  def  create_agent(**kwargs)  ->  Agent:

    """Demonstrate OpenAI Realtime with GitHub MCP server integration."""



    # Get GitHub PAT from environment

    github_pat  =  os.getenv("GITHUB_PAT")

    if  not  github_pat:

        logger.error("GITHUB_PAT environment variable not found!")

        logger.error("Please set GITHUB_PAT in your .env file or environment")

        raise  ValueError("GITHUB_PAT environment variable not found")



    # Check OpenAI API key from environment

    openai_api_key  =  os.getenv("OPENAI_API_KEY")

    if  not  openai_api_key:

        logger.error("OPENAI_API_KEY environment variable not found!")

        logger.error("Please set OPENAI_API_KEY in your .env file or environment")

        raise  ValueError("OPENAI_API_KEY environment variable not found")



    # Create GitHub MCP server

    github_server  =  MCPServerRemote(

        url="https://api.githubcopilot.com/mcp/",

        headers={"Authorization":  f"Bearer {github_pat}"},

        timeout=10.0, # Shorter connection timeout

        session_timeout=300.0,

    )



    # Create OpenAI Realtime LLM (uses OPENAI_API_KEY from environment)

    llm  =  Realtime(model="gpt-4o-realtime-preview-2024-12-17")



    # Create real edge transport and agent user

    edge  =  getstream.Edge()

    agent_user  =  User(name="GitHub AI Assistant",  id="github-agent")



    # Create agent with GitHub MCP server and Gemini Realtime LLM

    agent  =  Agent(

        edge=edge,

        llm=llm,

        agent_user=agent_user,

        instructions="You are a helpful AI assistant with access to GitHub via MCP server. You can help with GitHub operations like creating issues, managing pull requests, searching repositories, and more. Keep responses conversational and helpful. When you need to perform GitHub operations, use the available MCP tools.",

        processors=[],

        mcp_servers=[github_server],

    )

    logger.info("Agent created with OpenAI Realtime and GitHub MCP server")

    logger.info(f"GitHub server: {github_server}")



    return  agent



async  def  join_call(agent:  Agent,  call_type:  str,  call_id:  str,  **kwargs)  ->  None:

    try:

        # Set up event handler for when participants join

        @agent.subscribe

        async  def  on_participant_joined(event:  CallSessionParticipantJoinedEvent):

            # Check MCP tools after connection

            available_functions  =  agent.llm.get_available_functions()

            mcp_functions  =  [

                f  for  f  in  available_functions  if  f["name"].startswith("mcp_")

            ]

            logger.info(

                f"✅ Found {len(mcp_functions)} MCP tools available for function calling"

            )

            await  agent.say(

                f"Hello {event.participant.user.name}! I'm your GitHub AI assistant powered by OpenAI Realtime. I have access to {len(mcp_functions)} GitHub tools and can help you with repositories, issues, pull requests, and more through voice commands!"

            )



        # ensure the agent user is created

        await  agent.create_user()



        # Create a call

        call  =  await  agent.create_call(call_type,  call_id)



        # Have the agent join the call/room

        logger.info("🎤 Agent joining call...")

        with  await  agent.join(call):

            logger.info(

                "✅ Agent is now live with OpenAI Realtime! You can talk to it in the browser."

            )

            logger.info("Try asking:")

            logger.info("  - 'What repositories do I have?'")

            logger.info("  - 'Create a new issue in my repository'")

            logger.info("  - 'Search for issues with the label bug'")

            logger.info("  - 'Show me recent pull requests'")

            logger.info("")

            logger.info(

                "The agent will use OpenAI Realtime's real-time function calling to interact with GitHub!"

            )



            # Run until the call ends

            await  agent.finish()

    except  Exception  as  e:

        logger.error(f"Error with OpenAI Realtime GitHub MCP demo: {e}")

        logger.error("Make sure your GITHUB_PAT and OPENAI_API_KEY are valid")

        import  traceback

        traceback.print_exc()



    # Clean up

    await  agent.close()

    logger.info("Demo completed!")

if  __name__  ==  "__main__":

    cli(AgentLauncher(create_agent=create_agent,  join_call=join_call))

export  OPENAI_API_KEY=sk-...

export  STREAM_API_KEY=...

export  STREAM_API_SECRET=...

export  GITHUB_ACCESS_TOKEN=ghp_...



uv  run  main.py