How to setup and run locally with Ollama?

[playground]

I just stumbled into this and it looks very interesting.

I have a MCP server running locally with MCP host and Ollama. Is there an example on how to setup mcp-framework running locally with Ollama via http?

[aman-11]

check this out https://github.com/modelcontextprotocol/typescript-sdk#backwards-compatibility. this would help you assuming you have a mcp server running on any port.

I would recommend using StreamableHTTPServerTransport since SSE is deprecated now.
This is smth i could code to test it out. I read few docs around and came with this client setup. Let me know if this helps. Though I will try this tmrw since even I need to use this http setup

  const model = new ChatOllama({ model: "qwen2.5-coder:7b" });
      const mcpClient = new Client({
         name: 'streamable-http-client',
         version: '1.0.0'
       });
       const transport = new StreamableHTTPClientTransport(
         new URL('http://localhost:8070') // the http url your server is running on, default it would be 8070
       );
       await mcpClient.connect(transport);
       console.log("Connected using Streamable HTTP transport");
       const tools = await this.mcp.listTools(); // use when chat models are not from langchain SDK's
       const tools = await loadMcpTools('weather-mcp-serve', this.mcp); // use when chat models are not from langchain SDK's since it would transform the tools typings to respect langchain SDK
      model.bindTools(tools); // to bind directly 
      // OR bind the tools with llm
      this.agent = createReactAgent({
        llm: this.model,
        tools: tools,
      });

[QuantGeekDev]

Hey @playground — great question!

mcp-framework is a server-side framework, so it handles building and running MCP servers (exposing tools, resources, and prompts over stdio, SSE, or HTTP Stream). How a client connects to that server is outside our scope, but I can point you in the right direction.

Ollama doesn't natively support MCP yet (ollama/ollama#7865), so you'll need a bridge layer. A few options:

• mcp-client-for-ollama — A TUI client that connects Ollama to MCP servers, supports stdio, SSE, and Streamable HTTP transports
• ollama-mcp-bridge — A Python API layer that sits in front of Ollama and injects MCP tools into chat requests
• LangChain — As @aman-11 mentioned above, you can use ChatOllama + StreamableHTTPClientTransport from the MCP SDK

On our side, you'd set up an HTTP Stream server like this:

import { MCPServer } from "mcp-framework";

const server = new MCPServer({
  transport: {
    type: "httpStream",
    options: {
      port: 8070,
    },
  },
});

server.start();

Then any of the bridges above can connect to http://localhost:8070/mcp. Check out our HTTP Quickstart for the full setup.

Hope that helps!