Key facts
| AI SDK compatibility | OpenAI-compatible endpoint works with the Vercel AI SDK |
| Models | 30+ models in one catalogue, from free tiers to frontier reasoning |
| Pricing | Flat monthly self-serve plans with unlimited fair-use usage |
| Free tier | plugsky-micro and plugsky-lite, no card required |
| Trial | 14-day full-access trial for stronger models |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped |
| Relationship | Provider underneath the gateway, not a gateway replacement |
TL;DR
- The gateway gives Vercel apps one endpoint across providers.
- Plugsky supplies the models underneath through an OpenAI-compatible API.
- The Vercel AI SDK accepts OpenAI-compatible providers directly.
- Keep the gateway for observability and fallbacks if your app benefits.
- Flat monthly pricing makes the provider side predictable.
How it works, step by step
- Decide whether the gateway's routing and observability earn their place in your stack.
- Create a Plugsky key on the free plan and point a staging AI SDK client at the endpoint.
- If you keep the gateway, register Plugsky as an OpenAI-compatible upstream.
- Verify streaming, tool calls and structured output through your chosen path.
- Compare latency and error handling with and without the gateway hop.
- Choose the simplest topology that meets your reliability requirements.
Try it yourself
Open the Vercel AI Gateway cost calculator →
What the gateway solves for Vercel apps
Vercel AI Gateway provides a single endpoint across model providers with routing, fallbacks and usage visibility, and it integrates with the AI SDK that Vercel developers already use. For front-end teams shipping AI features quickly, that removes a lot of glue code.
It is still a gateway. It does not own model quality, residency or pricing shape; those belong to the provider underneath. Choosing the provider is therefore the more consequential decision, and it is independent of whether you keep the gateway.
Plugsky as the provider underneath
Plugsky exposes an OpenAI-compatible endpoint, which the Vercel AI SDK can consume like any OpenAI-style provider. You get 30+ models, live streaming, JSON mode, function calling and embeddings, plus deployment options that gateways cannot offer on their own: customer VPC, on-prem and air-gapped environments.
- Use one key for chat, embeddings and agents.
- Keep the AI SDK client code and change configuration.
- Test streaming and tool calls through the full path.
- Measure the extra hop if you keep the gateway.
- Keep configuration in environment variables so rollback is instant.
Topology decisions and honest limits
Two questions decide the topology. First, do you need multi-provider routing or customer-managed keys? If yes, keep a gateway. Second, do you need private deployment or residency guarantees? If yes, choose a provider that offers them rather than assuming the gateway will.
Plugsky does not replace gateway features such as cross-provider fallbacks and detailed request tracing. Audio, images, moderation, files, batch, fine-tuning, assistants and responses are coming soon; chat, streaming, tools, embeddings, RAG and agents are live. See the live pricing page for current plans.
Honest comparison
| Concern | Plugsky | Vercel AI Gateway | Direct provider setup |
|---|---|---|---|
| Role | Inference provider | Gateway for Vercel apps | Provider client per app |
| AI SDK fit | OpenAI-compatible endpoint | Native AI SDK integration | Works with AI SDK too |
| Pricing shape | Flat monthly self-serve | Separate gateway plans | Mixed provider bills |
| Deployment | Cloud, VPC, on-prem, air-gapped | Vercel-managed | Provider-dependent |
| Fallbacks | Configure in your app | Built into the gateway | You implement them |
Frequently asked questions
Does Plugsky replace Vercel AI Gateway?
No. The gateway routes and observes traffic; Plugsky is the provider that serves models. They work together, with Plugsky registered as an OpenAI-compatible upstream.
Does the Vercel AI SDK work with Plugsky?
Yes. The AI SDK supports OpenAI-compatible providers, so you configure the base URL and key and keep your client code.
Is there a free plan?
Yes. plugsky-micro and plugsky-lite are free with no credit card, and a 14-day full-access trial opens stronger models.
How is pricing structured?
Self-serve plans are flat monthly with unlimited fair-use usage and no per-token billing. See the live pricing page for current plans.
Do I need the gateway if I use one provider?
Not necessarily. A single provider plus built-in analytics is simpler; keep the gateway if you need cross-provider fallbacks or centralised key management.
Can I deploy Plugsky privately?
Yes. VPC, on-prem and air-gapped deployments are available, which is beyond what a managed gateway alone can provide.