Pura for developers
Integrate Pura models into your product with OAuth, official SDKs and shared account usage.
First integration
Register, connect, invoke.
OAuth and identity
Consent and per-app grants.
API reference
Fields and authentication.
How it works
The user connects their Pura account to your application and approves its scopes. Your application receives OAuth tokens and sends inference to Pura. Pura selects the internal account key, checks the model and limits, forwards to the provider and records usage.
The gateway key and provider credentials stay on Pura’s server. Each application has a separate grant; all authorized apps for an account share the same limits.
Choose your integration path
For a SaaS backend, register a confidential app and handle exchange and token storage on your server. For a public client, use the Connect button with PKCE and no client secret. In both cases, the user authorizes a specific app.
Subscriptions use delegated inference endpoints. API keys and the pay-as-you-go wallet use a separate path, documented in the Pura Docs gateway guide.
Contract and availability
SDK and webhook contract: 2026-10-01. The proprietary SDK packages use version 0.1.0. For a published release, install @pura-ai/sdk from npm or pura-llm-sdk from PyPI; use repository artifacts for source development. This documentation describes the implemented contract; operational availability depends on the deployed release.
capabilities() reports the services available to the account without exposing model identifiers. Pura configures one standard model per service in its own environment variables. Audio requires a verified deployment and plan activation.
Manual map
Start with the quickstart and application registration. Read OAuth and token storage before implementing chat or audio. Use the API reference for request fields and the webhook guide for lifecycle updates.