Vercel AI SDK Social Media: Posta as a Tool

Why Posta + Vercel AI SDK?

Path 1 — Posta MCP via experimental_createMCPClient

The AI SDK exposes an experimental MCP client that loads any MCP server's tools and hands them to generateText or streamText.

// app/api/post/route.ts
import { anthropic } from '@ai-sdk/anthropic'
import { experimental_createMCPClient as createMCPClient, generateText } from 'ai'
import { Experimental_StdioMCPTransport as StdioTransport } from 'ai/mcp-stdio'

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

  const mcp = await createMCPClient({
    transport: new StdioTransport({
      command: 'npx',
      args: ['-y', 'posta-mcp'],
      env: { POSTA_API_TOKEN: process.env.POSTA_API_TOKEN! },
    }),
  })
  const tools = await mcp.tools()

  try {
    const result = await generateText({
      model: anthropic('claude-sonnet-4-6'),
      tools,
      prompt,
    })
    return Response.json({ text: result.text })
  } finally {
    await mcp.close()
  }
}

Path 2 — Define a typed tool() around the REST API

Simpler when you want one deterministic call exposed to the model. Zod gives the AI SDK a typed schema; the model knows exactly what shape to call.

import { openai } from '@ai-sdk/openai'
import { generateText, tool } from 'ai'
import { z } from 'zod'

const POSTA = 'https://api.getposta.app'
const HEADERS = {
  'Authorization': `Bearer ${process.env.POSTA_API_TOKEN}`,
  'Content-Type': 'application/json',
}

const schedulePost = tool({
  description: 'Create a Posta draft on the given social accounts and schedule it.',
  parameters: z.object({
    caption: z.string().describe('Post caption'),
    socialAccountIds: z.array(z.number()).describe('Posta social_account IDs to publish to'),
    scheduledAt: z.string().describe('ISO 8601 timestamp'),
    mediaIds: z.array(z.string()).optional().describe('Optional Posta media IDs'),
  }),
  execute: async ({ caption, socialAccountIds, scheduledAt, mediaIds }) => {
    // Step 1 — create the draft
    const createRes = await fetch(`${POSTA}/v1/posts`, {
      method: 'POST',
      headers: HEADERS,
      body: JSON.stringify({
        caption,
        socialAccountIds,
        ...(mediaIds?.length ? { mediaIds } : {}),
      }),
    })
    if (!createRes.ok) throw new Error(`Posta create ${createRes.status}`)
    const { id } = await createRes.json()

    // Step 2 — schedule it
    const scheduleRes = await fetch(`${POSTA}/v1/posts/${id}/schedule`, {
      method: 'POST',
      headers: HEADERS,
      body: JSON.stringify({ scheduledAt }),
    })
    if (!scheduleRes.ok) throw new Error(`Posta schedule ${scheduleRes.status}`)
    return { id, scheduledAt }
  },
})

const result = await generateText({
  model: openai('gpt-4o'),
  tools: { schedulePost },
  prompt: 'Schedule a LinkedIn post about our v2 launch for tomorrow 9am CET.',
})

Common patterns