Key facts
| Relationship | Plugsky is the inference API; Portkey can route and observe it |
| Models | 38 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 |
| Built-in analytics | Usage analytics are available in the dashboard |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped |
TL;DR
- Portkey governs traffic; Plugsky serves models.
- They compose because Plugsky is OpenAI-compatible.
- Keep gateway guardrails and caching; add Plugsky as an upstream.
- Plugsky's built-in analytics cover basic monitoring needs.
- Consolidate provider keys where the gateway is not required.
How it works, step by step
- Separate your needs: model inference, gateway policy, or both.
- For inference, create a Plugsky key on the free plan and test a workload.
- If you keep Portkey, register Plugsky as an OpenAI-compatible upstream.
- Verify guardrails and caching behave the same with the new upstream.
- Compare built-in analytics with your gateway dashboards.
- Document which traffic uses gateway features and which calls Plugsky directly.
Try it yourself
Open the Portkey API cost calculator →
What Portkey provides
Portkey sits in front of model providers and applies policy: routing rules, fallbacks, caching, guardrails, budget limits and observability. For teams with multiple providers, compliance requirements or a need to centralise keys, a gateway is a legitimate architectural layer.
It is not an inference provider. That distinction keeps the evaluation honest: if your problem is policy and visibility, keep a gateway; if your problem is where models run and what they cost, choose the provider underneath.
Gateway and provider, composed
Plugsky exposes an OpenAI-compatible endpoint, so it can be registered as an upstream in Portkey like any OpenAI-style provider. You keep guardrails and caching, and you gain a catalogue of 38 models, flat monthly self-serve pricing and sovereign deployment options where required.
- Add Plugsky as an upstream and canary a share of traffic.
- Confirm streaming and tool payloads flow through unchanged.
- Reconcile spend between gateway and provider dashboards.
- Scope keys per environment to simplify rotation.
- Test gateway retries against provider timeouts before production rollout.
When built-in analytics are enough
Plugsky includes usage analytics in the dashboard: volume, model mix and errors. For single-provider stacks without strict policy needs, that covers day-to-day monitoring. Gateways earn their place when you need custom guardrails, semantic caching, complex routing or customer-managed keys.
Plugsky is not a full gateway and does not claim to replace one, but built-in analytics cover everyday monitoring. Keep one source of truth for budgets; reconciling two dashboards wastes time. 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 | Portkey | Typical custom middleware |
|---|---|---|---|
| Role | Inference API for 38 models | AI gateway and observability | You build routing and policy |
| API style | OpenAI-compatible | Gateway in front of providers | Varies |
| Pricing shape | Flat monthly self-serve | Separate gateway plans | Engineering time |
| Policy features | Keys and usage analytics | Guardrails, caching, budgets | Whatever you implement |
| Deployment | Cloud, VPC, on-prem, air-gapped | Hosted or self-hosted | Varies |
Frequently asked questions
Does Plugsky replace Portkey?
No. Portkey is a gateway with routing, caching and guardrails; Plugsky is the inference provider. They typically run together, with Plugsky as an upstream.
Can Portkey route to Plugsky?
Yes. Plugsky is OpenAI-compatible, so it can be added as an upstream provider in the gateway.
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 still need a gateway for observability?
Only if you need deep tracing, custom guardrails or customer-managed keys. Plugsky provides built-in usage analytics for everyday monitoring.
Can Plugsky deploy privately behind a gateway?
Yes. VPC, on-prem and air-gapped deployments are available, which keeps gateway traffic inside your network.