does it work with ollama? are there any code examples?

[fahadshery]

Hi,

As the title suggests, does it work with Ollama? Is there an example?

Cheers,

[moonbox3]

Hi @fahadshery, are you looking for Python or .NET samples?

[rakeshreddy91]

I’m interested in setting up Ollama, or using a mix of offline and online models based on the query and MCP tool integration.

[evan-caswell]

@fahadshery @rakeshreddy91

Here is a simple implementation of using Ollama with MCP tools, assuming you are using Python.

To use Ollama, access its OpenAI compatible endpoint using OpenAIChatClient.
I used Docker MCP gateway to expose MCP servers.

import asyncio
import os

from agent_framework import (
    ChatAgent,
    ChatMessage,
    MCPStreamableHTTPTool,
    Role,
    TextContent,
)
from agent_framework.openai import OpenAIChatClient
from dotenv import load_dotenv

load_dotenv()


async def main() -> None:
    async with MCPStreamableHTTPTool(
        name="docker_mcp", url="http://localhost:8080/mcp"
    ) as mcp_server:
        # You can specify the tools you want exposed by the docker mcp gateway.
        mcp_server.allowed_tools = ["convert_time", "get_current_time"]

        chat_client = OpenAIChatClient(
            model_id="qwen3:4b",
            # Ollama OpenAI compatible endpoint.
            base_url="http://localhost:11434/v1/",
            # Key from .env file for online models.
            api_key=os.getenv("OLLAMA_API_KEY", ""),
        )

        agent = ChatAgent(
            chat_client=chat_client,
            instructions="You are a helpful assistant with access to the current time.",
            name="Assistant",
            tools=mcp_server,  # Pass mcp_server as tools.
        )

        user_message = ChatMessage(
            Role.USER,
            contents=[TextContent("What time is it in New York City, New York?")],
        )
        response = await agent.run(user_message)

        print(response.text)


if __name__ == "__main__":
    asyncio.run(main())