Key facts
| Tools shape | OpenAI-compatible tools array on /v1/chat/completions with tool_calls responses (live) |
| Free plan | Two free AI models — plugsky-micro and plugsky-lite — with no card required |
| Trial | 14-day full-access trial to test paid-tier models and features |
| Frameworks | Works with the OpenAI SDKs, LangChain, LlamaIndex and the Vercel AI SDK |
| Cost model | Flat self-serve plans with unlimited fair-use usage; no per-token charges |
| Model choice | 30+ models behind one endpoint, from free tiers to frontier reasoning |
| Migration | OpenAI-compatible, so switching back or to another provider is a base URL change |
| Roadmap | Assistants and responses endpoints are coming soon; the chat completions tool loop is live |
TL;DR
- Ship your first tool loop on the free plan without a credit card.
- Keep three to five well-described tools; add more only when selection stays reliable.
- Use flat pricing so a viral week does not become a billing emergency.
- Keep your SDK code portable — the API stays OpenAI-compatible.
- Upgrade to paid tiers or the 14-day full-access trial when usage justifies it.
How it works, step by step
- Create an account, generate an API key on the free plan and point the OpenAI SDK at the Plugsky base URL.
- Define three to five tools that map to real product actions — create record, fetch status, send notification.
- Implement the loop server-side and return tool errors to the model so it can recover.
- Test with the function calling tester and a small set of prompts your users will actually send.
- Add streaming so the interface feels responsive while tools execute.
- Watch token telemetry and error rates, then move heavy paths to a paid tier or the 14-day full-access trial.
Original data
Try it yourself
Open the function calling tester →
From idea to first tool loop in an afternoon
You do not need orchestration infrastructure to ship a useful agent. Take your existing OpenAI SDK call, change the base URL, and add a tools array with the two or three actions your product actually needs. The model returns tool_calls, your endpoint executes them, and you append the results as role=tool messages until the assistant answers.
Start on plugsky-micro or plugsky-lite, the two free models on the free plan. They are strong enough for classification, extraction and routing, which is most of what an early agent does. Keep the tool count low: a tight, well-described set beats twenty overlapping functions that confuse model selection.
Cost and capacity before revenue
Early traction is exactly when per-token billing hurts. A demo day spike or a crawling bot can turn into a bill you cannot pay. Plugsky self-serve plans are flat monthly with unlimited fair-use usage, so growth changes your request rate, not your invoice shape; token counts still return in each response for your own dashboards.
The free plan carries no card requirement, and the 14-day full-access trial lets you test paid-tier models before committing. That sequencing — validate on free models, trial the paid tier, then choose a plan — keeps burn low while you find product-market fit.
Keep the exit open
The OpenAI-compatible shape is your insurance policy. If you ever need to move providers or run a hybrid setup, you change the base URL and model names, not your application code. That portability also means frameworks you already use — LangChain, LlamaIndex, the Vercel AI SDK — keep working.
As you grow, the same endpoint gives you 30+ models. You can move summarisation to a cheap model, reasoning to a frontier one, and add routing or fusion later without re-architecting. Build the loop you need this week and let the platform grow underneath it.
Honest comparison
| Concern | Plugsky | Typical per-token API | Self-hosting open models |
|---|---|---|---|
| Upfront cost | Free plan with two models, no card | Prepaid credits | GPU reservation or spot |
| Cost shape | Flat monthly, unlimited fair use | Scales with tokens | Fixed GPU, variable ops |
| Time to first call | Minutes with the OpenAI SDK | Minutes | Days to weeks |
| Model choice | 30+ models, one endpoint | Provider catalogue | Whatever fits your GPUs |
| Engineering load | API integration only | API integration only | Serving, scaling, monitoring |
Frequently asked questions
Is there really a free plan?
Yes. The free plan includes two free AI models, plugsky-micro and plugsky-lite, with API keys and no credit card required.
What is the trial?
A 14-day full-access trial lets you test paid-tier models and features. When it ends you pick a plan; there is no automatic charge.
Can I keep using the OpenAI SDK?
Yes. Plugsky exposes an OpenAI-compatible API, so you change the base URL and model name and keep your existing code.
How many tools should a first version have?
Three to five. Cover the core actions, write clear descriptions and typed fields, and expand only when the model still selects correctly.
Will costs spike if we go viral?
Self-serve plans are flat monthly with unlimited fair-use usage, so usage changes your request rate, not your invoice shape. See the live pricing page for current plans.
What if we outgrow the free models?
Move heavier or higher-stakes paths to paid tiers, or start the 14-day full-access trial to benchmark before choosing a plan.
Can we self-host later?
Yes. The same API runs in your VPC, on-prem or air-gapped when your customers or regulators require it.