For institutions
Governance that scales from one course to a whole campus
Sentinel Hub is multi-tenant from the ground up: every school, department, class, and lab gets its own isolated users, models, devices, limits, and reports. Start with a single course. Grow by adding tenants — not by renegotiating anything.
Office by office
Built for the offices that have to say yes
- Procurement
- One gateway account replaces a shifting roster of per-student subscriptions and personal-card reimbursements. Budgets live where they belong — as server-enforced quotas on tenants and keys, with cost estimates recorded per request — so spending questions are answered by a report, not a shoebox of receipts.
- Information technology
- Students work through a browser dashboard and the official WireGuard client — the industry-standard VPN protocol, not custom software. The private AI API is unreachable from the public internet, internal services stay unexposed, and provider credentials are encrypted and never touch a student machine. Suspending a key, device, or tenant takes effect across the API promptly.
- Academic oversight
- Administrators see who is using AI, which models, and how much capacity — filterable by date, model, key, and device. Logging is metadata-only by default: model, tokens, cost, status, request ID. Prompts and responses are not retained unless your institution explicitly enables and discloses that policy. An acceptable-use framework covers academic integrity, privacy, and prohibited uses.
The alternative
Sentinel Hub vs. enterprise chat licensing
Enterprise chat licenses are easy to buy and easy to under-use: students already have free chat models, so a seat license largely duplicates what the classroom already has. A gateway funds what coursework actually demands — API access students can build against, on models the institution chooses.
| Capability | Enterprise chat license | Sentinel Hub gateway |
|---|---|---|
| Chat with frontier models | Included | Included — full API access |
| Build real applications against an API | Not included | OpenAI-compatible endpoints |
| Multiple providers behind one interface | Single vendor | Routes to any approved provider |
| Per-student and per-app limits | Seat-based, not metered | Quotas on keys, devices, and tenants |
| Local and open models | Not included | Ollama-hosted models, same API |
| Usage attributable to course and project | Aggregate seats | Per-key, per-device reporting |
| Pay only for what’s used | License expires | Metered per request |
Trust, stated plainly
What we promise every campus
Provider credentials are encrypted and never shown to students or third-party applications — students receive Sentinel keys, not provider keys. Every application key and every enrolled device is separately revocable. Tenants cannot see each other’s data, models, devices, keys, or usage records.
Every request is checked for tenant, key, device, model, quota, and policy before it reaches a provider. Errors come back sanitized, with a request ID that lets support trace the problem without ever asking a student for a key or a prompt. Secure behavior is the easiest behavior — by design.