feat: add @plasmate/ai — Vercel AI SDK integration

[dbhurley]

Summary

Adds integrations/vercel-ai/ — a new @plasmate/ai package that makes Plasmate's browser tools available as first-class tools in the Vercel AI SDK.

What it does

createPlasmateTools() spawns plasmate mcp as a stdio subprocess, connects via the AI SDK's experimental_createMCPClient + Experimental_StdioMCPTransport, and returns CoreTool objects ready for generateText / streamText.

import { createPlasmateTools } from '@plasmate/ai'
import { generateText } from 'ai'
import { openai } from '@ai-sdk/openai'

const { tools, close } = await createPlasmateTools()

const { text } = await generateText({
  model: openai('gpt-4o'),
  tools,
  maxSteps: 5,
  prompt: 'Summarize the top stories on news.ycombinator.com',
})

await close()

Package details

• Name: @plasmate/ai
• Version: 0.1.0
• License: Apache-2.0
• Peer deps: ai >= 4.0.0, @ai-sdk/mcp >= 0.0.1
• Build: dual ESM + CJS via tsup, full TypeScript types

Follows existing patterns

Structured like integrations/browser-use and integrations/langchain — lives under integrations/ alongside the other framework adapters.

Why this matters

The Vercel AI SDK is the most widely adopted TypeScript AI framework. This gives every Next.js / Node.js developer using the AI SDK a one-liner to drop Plasmate's browser capabilities into their agents — no manual MCP wiring required.

[vercel]

@dbhurley is attempting to deploy a commit to the dbh ventures Team on Vercel.

A member of the Team first needs to authorize it.