Frequently asked questions

A governance layer plus an OpenAI-compatible API in front of nine LLM providers. Your AI agent points at PGAIN instead of a provider; PGAIN enforces cost caps, audit logging, and skill injection on every call, then forwards to the provider using your own key.

OpenRouter is a routing marketplace you pay through. PGAIN is a governance harness: you bring your OWN provider keys (including OpenRouter), and PGAIN adds per-user cost caps, server-side audit, Bayesian model election, and injectable skills. You pay providers directly at cost — PGAIN takes no markup.

Yes. PGAIN is BYOK (bring your own key). Add a key for any of the 9 providers in Settings; it's encrypted at rest and used only for your runs. Ollama needs no key (it runs locally).

Every account has a daily spend cap (default $5/day). PGAIN meters the real provider cost of each call server-side; once you hit the cap, further calls return HTTP 402 until the next day. You can watch today's spend on your dashboard.

/v1/* is the OpenAI-compatible surface (chat/completions, embeddings, models) for drop-in SDK use. /api/prompt is PGAIN's higher-level compound endpoint that runs the architect+peer harness (planning, model election, skills) for multi-step tasks.

Skills are curated instruction packs (e.g. Karpathy guidelines, bug-diagnosis, code-review). Enable them on the Skills page; enabled skills are auto-injected into your agent's planning prompt on every call. You can also bring your own as signed manifests.

Your provider keys are Fernet-encrypted at rest and never logged or shared. Prompts and responses aren't recorded by default (an opt-in recorder exists for your own debugging). See the Privacy page.

Anything that speaks the OpenAI API: Hermes, Cline, Aider, Cursor, Continue, and raw OpenAI SDKs (Python/JS/curl). See Agents and Docs.

A Q3 2026 marketplace for premium skills, federated nodes, and custom plugins, with priority support. The core API stays free forever. Opt in for launch updates on the Pricing page.