vercel

Neo — Vercel AI SDK Integration

Use Neo's MCP server as a set of tools inside any Vercel AI SDK application. Neo executes AI/ML workloads locally on the user's machine via a daemon — files are written directly to their workspace, never to a remote server.

MCP server: https://mcpserver.heyneo.com/mcp Auth: Authorization: Bearer sk-v1-YOUR_KEY

Option A: MCP Client (recommended — zero tool duplication)

Connect to the hosted server with the Vercel AI SDK's built-in MCP client. All 7 Neo tools are loaded automatically — no manual schema definitions needed.

// lib/neo.ts
import { experimental_createMCPClient as createMCPClient } from 'ai';
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js';

export async function createNeoClient() {
  return createMCPClient({
    transport: new StreamableHTTPClientTransport(
      new URL('https://mcpserver.heyneo.com/mcp'),
      {
        requestInit: {
          headers: {
            Authorization: `Bearer ${process.env.NEO_SECRET_KEY}`,
          },
        },
      }
    ),
  });
}

// app/api/chat/route.ts  (Next.js App Router)
import { streamText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { createNeoClient } from '@/lib/neo';

export async function POST(req: Request) {
  const { messages } = await req.json();

  const neoClient = await createNeoClient();
  const neoTools = await neoClient.tools();

  const result = streamText({
    model: openai('gpt-4o'),
    tools: neoTools,
    messages,
    system: `You are an AI assistant with access to Neo, a local AI/ML execution backend.
Use neo_submit_task for any ML/AI work (training, fine-tuning, RAG, agents, data preprocessing).
Files are written directly to the user's local machine — never to a remote server.
After submitting, poll neo_task_status until COMPLETED, then call neo_get_messages for the final output.`,
    onFinish: async () => {
      await neoClient.close();
    },
  });

  return result.toDataStreamResponse();
}