MITPO relies on third-party providers to deliver parts of the product. This page describes those providers at a category level — what they do for us, and where processing typically occurs. Some categories include multiple providers that may be selected per request (for example, AI model providers differ by the feature you use).

If you need a specific data-flow diagram for a procurement process, email support@mitpo.com and we will provide one.

The application runs on a managed serverless platform for its web tier and edge delivery. Background jobs (scheduled crons, long-running generations) use providers suitable for each task.

• Vercel — web hosting, edge, build pipeline (US + global edge).
• Upstash QStash — scheduled task execution and webhook fan-out.
• Upstash Redis — rate limiting and short-lived caches.

Customer data is stored in a managed Postgres instance with row-level security enforced at the database layer. Generated assets (images, videos, audio) live in a private object-storage bucket and are served through an authenticated stream endpoint that re-signs per request.

• Neon (Postgres) — primary relational database, US region.
• Cloudflare R2 — private object storage for generated assets.

Authentication is handled by Better Auth with email + password or OAuth single sign-on. OAuth connections are initiated through the identity provider’s standard flow; only the scopes required to operate the product are requested.

• Better Auth — session management and password hashing.
• Google, GitHub — optional OAuth single sign-on.

Subscriptions and credit purchases are processed by Stripe. MITPO does not store full card numbers — only Stripe customer IDs and non-sensitive metadata (last four digits, card brand) that Stripe returns.

• Stripe — card processing, subscription management, tax calculation.

MITPO orchestrates multiple AI providers behind a unified model picker. Each generation is routed to the provider you (or your plan default) selected. Prompts and the resulting assets are sent only to the provider handling that specific request — we do not fan out your data across every provider.

Provider choice affects where your request is processed. The picker surfaces the provider badge on every model so you can choose explicitly where applicable.

• OpenRouter — unified router for Gemini 3 Pro Image, DeepSeek, and select chat models.
• Google Cloud Vertex AI — Imagen, Veo 3.1, Gemini chat.
• fal.ai — FLUX family, Kling, Seedance, Luma, Pika video models.
• Replicate — Wan video, select long-tail models.
• Runware — MiniMax Hailuo, Kling Avatar.
• OpenAI — GPT Image, text-to-speech, select chat models.
• ElevenLabs — premium voices for speech generation.

Transactional email (signup verification, password resets, incident notifications) goes through a standard provider. Marketing email, if ever enabled, would be explicitly opt-in and handled separately.

• Resend — transactional email delivery.

Error tracking and performance telemetry use providers that receive stack traces, request metadata, and user identifiers but are configured to strip sensitive payloads before ingestion.

• Sentry — error tracking and performance monitoring.

When MITPO adds a new provider that processes customer data or changes a core provider in a way that affects data flow, this page is updated. If you maintain a vendor allow-list and need advance notice of changes, email support@mitpo.com — we can add you to an internal notification list.