Vercel AI Gateway is a clean way to ship AI inside a Vercel project. The trade-off is that everything gets pulled into Vercel's billing and Vercel's ecosystem, including your AI spend. Mesh API does the same provider routing, retries, and observability work, and runs anywhere your code does, with negotiated rates and a bill that doesn't surprise you.

Quick Migration

If you're using the Vercel AI SDK with the AI Gateway provider:

diff

- import { gateway } from "@ai-sdk/gateway";+ import { createOpenAI } from "@ai-sdk/openai";++ const mesh = createOpenAI({+   baseURL: "https://api.meshapi.ai/v1",+   apiKey: process.env.MESH_API_KEY,+ });

Then your generateText, streamText, and tool calls keep working, just point them at mesh("auto") or mesh("anthropic/claude-opus-4-7") instead of the gateway model.

Why teams switch from Vercel AI Gateway

Capability	Vercel AI Gateway	Mesh API
OpenAI-compatible endpoint	Indirect (through AI SDK)	Direct, works with any OpenAI SDK
Works outside Vercel hosting	Bundled with Vercel	Yes, any framework, any cloud
Billing	Through Vercel invoice	Standalone, pay-as-you-go, multi-currency
Platform fee	Yes	Zero
Smart routing across cost / latency / capability	Provider routing only	Live scoring on every dimension
Negotiated provider rates	Limited	Across every major provider
Team management outside Vercel teams	Tied to Vercel teams	Native RBAC, SSO, per-team budgets
Per-key spend limits	Limited	Built in
Audit log	Limited	Full audit log
Pre-built RAG and prompt workflows	Manual	Ready-made
Multi-cloud / on-prem clients	Awkward	First-class

Migration steps

1. Sign up for Mesh

Get a key from the dashboard.

2. Keep the Vercel AI SDK, switch the provider

The Vercel AI SDK is great. You don't have to leave it. Just point its OpenAI provider at Mesh:

typescript

import { createOpenAI } from "@ai-sdk/openai";import { generateText } from "ai"; const mesh = createOpenAI({  baseURL: "https://api.meshapi.ai/v1",  apiKey: process.env.MESH_API_KEY,}); const { text } = await generateText({  model: mesh("auto"),  // or "anthropic/claude-opus-4-7", "openai/gpt-4.1", etc.  prompt: "Explain Mesh API in one sentence.",});

Streaming:

typescript

import { streamText } from "ai"; const result = await streamText({  model: mesh("auto"),  prompt: "Write a haiku about gateways.",}); for await (const delta of result.textStream) {  process.stdout.write(delta);}

Tool calling, structured output, and multimodal inputs all work the same way as with the Vercel AI Gateway provider, you're just on a different base URL.

3. Move billing out of Vercel

Set up Mesh billing once, recharge in your currency, and you've decoupled AI spend from your hosting invoice. Most finance teams ask for this within a quarter anyway.

4. Extend to non-Vercel workloads

Cron jobs, background workers, mobile backends, internal tools that don't sit on Vercel, they all hit api.meshapi.ai/v1 with the same key. One bill, one dashboard, one audit log across your whole stack.

What you gain on Mesh

Portability. Your gateway lives at a URL, not inside a hosting platform.
Zero platform fee. What the model costs is what you pay.
Smart routing. Every request scored on cost, latency, capability, and context window. "auto" is auditable, not a black box.
Team controls. Per-key spend limits, RBAC, SSO, audit log, ZDR, outside any single hosting provider's account model.
Negotiated rates and multi-currency. Built for teams who care what the AI bill looks like in their actual currency.

Migrate from Vercel AI Gateway