Key facts
| API compatibility | OpenAI-compatible chat, embeddings and function calling |
| Use cases | Dossier and protocol drafting support, literature and label Q&A, adverse event intake triage |
| Human review | Keep a qualified reviewer in the loop for any GxP-relevant output |
| Data controls | API keys, RBAC, SSO, audit logs, region selection and retention settings |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped options |
| Models | 30+ models from free aliases to frontier reasoning |
| Pricing model | Flat monthly self-serve plans with unlimited fair-use usage |
| Free tier | plugsky-micro and plugsky-lite on the free plan, no card required |
TL;DR
- Use retrieval plus citations for label, protocol and literature questions.
- Keep a qualified human reviewer for anything that touches a controlled process.
- Audit logs, RBAC and region selection are the controls to configure first.
- VPC, on-prem and air-gapped options cover controlled and restricted environments.
- Start on the free plan, then evaluate frontier models on the same workload.
How it works, step by step
- Pick a bounded, reviewable workflow such as literature triage or protocol summarisation.
- Define the data boundary: which documents may be sent, where they are processed and how long logs are kept.
- Ground answers in your own corpus with embeddings and RAG, and require citations to the source passage.
- Build against the OpenAI-compatible endpoint so the application is portable across deployment modes.
- Add reviewer sign-off, versioned prompts and an evaluation set before any wider rollout.
- Document your own validation evidence for each workflow; the platform provides controls, not your compliance case.
Try it yourself
Open the AI data residency checklist →
Where an AI API fits in pharma workflows
The strongest early fits are text-heavy and review-gated:
- Regulatory writing: draft and restructure dossier sections, then route to a qualified reviewer.
- Literature and label Q&A: answer internal questions across publications, labels and SOPs with citations to the source passage.
- Adverse event intake: extract and normalise fields from reports for a pharmacovigilance reviewer to confirm.
- Clinical document summaries: condense protocols, amendments and investigator materials for internal briefings.
None of these should write directly into a controlled record without human confirmation.
Security, audit and data handling
Pharma buyers usually start with three questions: where does data go, who can see it, and what is logged? Plugsky supports scoped API keys, RBAC and SSO for people, audit logs for requests, and region selection for processing. Set retention deliberately, and keep personal or patient-identifiable data out of prompts unless a documented process and lawful basis exist. The platform offers the controls; your validation and quality system still own the evidence trail for each workflow.
Deployment options for controlled environments
Deployment choice follows the data: public cloud for non-sensitive R&D content, your VPC where network isolation and private connectivity are required, on-prem for strict control, and air-gapped where nothing may leave the environment. Because Plugsky is OpenAI-compatible, retrieval code, prompts and tests stay the same across all four. That makes it practical to validate a workflow in the cloud and move it to a stricter environment without re-platforming the application.
A pragmatic rollout
Start with plugsky-micro or plugsky-lite on the free plan against a small, low-risk corpus to prove retrieval and citation quality. Then use the 14-day full-access trial to compare frontier models on the same questions, scoring factual support, missing-citation rate and reviewer minutes. Keep prompts and evaluation sets versioned, pin the model you approve, and re-run the set whenever the corpus or model changes. Expand to a second workflow only after the first shows stable, reviewable output.
Honest comparison
| Capability | Plugsky | Typical per-token API | Building in-house |
|---|---|---|---|
| API compatibility | OpenAI-compatible chat, embeddings and tools | Usually compatible | Full rewrite |
| Audit and access | API keys, RBAC, SSO and audit logs | Varies by provider | You build all of it |
| Deployment | Cloud, VPC, on-prem and air-gapped | Mostly cloud-only | You operate GPUs and serving |
| Pricing | Flat monthly self-serve, fair-use usage | Per-token, harder to forecast | GPU plus operations cost |
| Model choice | 30+ models behind one API | Varies by provider | You host every model |
Frequently asked questions
Can we keep our existing OpenAI SDK code?
Yes. Plugsky is OpenAI-compatible: change the base URL and model name and keep your SDK, retries and prompt structure.
Is there a free plan?
Yes. The free plan gives you plugsky-micro and plugsky-lite with no credit card, which is enough to prototype retrieval and citation quality.
Does Plugsky make our workflow GxP validated?
No. Plugsky provides platform controls such as access, logging and region selection. Validation evidence and quality-system decisions remain yours.
Can we deploy in our own environment?
Yes. VPC, on-prem and air-gapped deployments are available, so you can match the environment to the sensitivity of the data.
How does pricing work?
Self-serve plans are flat monthly with unlimited fair-use usage. See the live pricing page for current plans and enterprise terms.
Which endpoints are live today?
Chat, streaming, JSON mode, function calling, embeddings, RAG and agents are live. Audio, images, moderation, files, batch, fine-tuning, assistants and responses are coming soon.
What should we evaluate first?
Citation support and reviewer-time savings on a small labelled set of real internal questions. Those two metrics predict whether the workflow is worth scaling.