Key facts
| ChatGPT agent features | Hosted inside the ChatGPT product, not the API surface |
| Model choice | OpenAI model family only |
| Plugsky interface | OpenAI-compatible chat completions: change base URL and model name |
| Model breadth | 30+ models behind one API key, from free tiers to frontier |
| Free tier | plugsky-micro and plugsky-lite free with no card required |
| Orchestration | Runs in your code with your tools, state and approval gates |
| Deployment | Cloud, VPC, on-prem and air-gapped with region choice |
| Honest limit | No consumer chat UI equivalent; you build the interface |
TL;DR
- Consumer agent features and an agent API solve different problems.
- An API-first option keeps orchestration, data and review in your code.
- OpenAI-compatible endpoints mean a base URL and model name change, not a rewrite.
- You give up the packaged chat interface; you gain control and portability.
- Check status labels: some OpenAI-style endpoints are still coming soon.
How it works, step by step
- List which ChatGPT agent behaviours you actually depend on and where they run today.
- Decide whether you need a chat UI or an agent embedded in your own product.
- Prototype the loop with direct API calls: tools, state, retries and logging.
- Point an OpenAI-format client at Plugsky and test the same prompts and tools.
- Compare tool-call reliability, latency and cost per completed task.
- Add guardrails: scoped keys, approval gates and usage limits before rollout.
- Keep the interface OpenAI-compatible so switching again stays cheap.
Try it yourself
Open the ChatGPT alternative finder →
Consumer agent versus agent API
ChatGPT's agent features are a product: a hosted interface where the assistant browses, runs tasks and returns results inside OpenAI's environment. That is convenient for individuals and small teams, and the wrong shape for product teams that need agents inside their own application, with their own data, permissions and audit trail.
An agent API is the other shape. You call a model endpoint, declare tools, keep state yourself and decide what requires human approval. There is no packaged UI, but there is also no ceiling on how the agent integrates with your systems.
- Hosted agent: fast to try, limited control, one vendor's models.
- Agent API: more engineering, full control, portable across providers.
- Hybrid: a hosted tool for internal experiments, an API for the product path.
What changes when you move to an API-first stack
The code is familiar if you already use OpenAI-format clients. You change the base URL, map model names to the catalogue, and keep your SDK calls. What you gain: access to 30+ models on one key, so routine steps can run on small fast models and hard steps on frontier models without a second integration. What you own: orchestration, memory, evaluation and the user interface.
Plan for the operational work honestly. You will choose a vector store for memory, decide how long traces are retained, and build the approval flow for risky tools. None of that is difficult, but none of it is included in a chat subscription either.
Where Plugsky fits and where it does not
Plugsky is an API platform, not a chat product. It serves 30+ models through an OpenAI-compatible endpoint with live function calling, streaming, JSON mode, embeddings, RAG and agents, plus scoped keys, RBAC, SSO/SCIM and audit logs for production use, and deployment from shared cloud to VPC, on-prem and air-gapped.
It does not give you a hosted agent UI, browsing infrastructure or consumer apps, and some OpenAI-style endpoints — assistants, responses, batch, files, fine-tuning and audio — are coming soon rather than live. If your need is a managed assistant product, a hosted tool is the better fit; if your need is an agent inside your own software, an API platform is. Self-serve plans are flat monthly and the free tier covers plugsky-micro and plugsky-lite; current plans are on the live pricing page.
Honest comparison
| Dimension | ChatGPT agent features | Plugsky API | Building in-house |
|---|---|---|---|
| Shape | Hosted product with chat UI | API and runtime, no UI | You build everything |
| Models | OpenAI family | 30+ models, one key | Whatever you host |
| Control | Limited to product settings | Your tools, state and gates | Total |
| Deployment | Vendor cloud | Cloud, VPC, on-prem, air-gapped | Your infrastructure |
| Effort | Lowest | Moderate | Highest |
Frequently asked questions
Can I use my OpenAI SDK with Plugsky?
Yes. Plugsky exposes an OpenAI-compatible chat completions endpoint, so you change the base URL and model names and keep your existing SDK code and tests.
Do I get a chat interface like ChatGPT?
No. Plugsky is an API platform. You build or buy the interface, which is what gives you control over the agent's behaviour, data and approvals.
Is there a free way to test it?
Yes. The free plan includes plugsky-micro and plugsky-lite with no card required, and a 14-day full-access trial is available for evaluating the full catalogue.
What happens to browsing and computer use?
Those are tools you implement, typically with a headless browser in your own sandbox. Plugsky provides function calling so the agent can invoke them.
How is pricing structured?
Self-serve plans are flat monthly with unlimited fair-use usage rather than per-token billing. See the live pricing page for current plans.
Can I deploy in my own environment?
Yes. Plugsky supports region choice, VPC, on-prem and air-gapped deployments for teams with residency or export-control requirements.
Which ChatGPT-era workloads should stay put?
Anything that depends on OpenAI-hosted assistants, file search or audio endpoints, which are coming soon on Plugsky. Keep those running where they work today.