pevier

Inspiration;

AI agents can create content instantly, but giving them unrestricted access to public accounts is risky. I built Pevier to place enforceable policy, human approval and verifiable evidence between an autonomous agent and the publish button.

What it does; Pevier evaluates every publishing request and returns ALLOW, HOLD or BLOCK. It supports safe dry runs, explicit confirmation for live posts, emergency publishing stops and a SHA-256 hash-linked audit trail. Instagram Reels and Bluesky posts are currently supported.

How I built it; I used Next.js, React, TypeScript, Prisma and PostgreSQL. Google OAuth handles user authentication, while Instagram and Bluesky OAuth let users connect their own accounts securely. Vercel Blob temporarily transfers Reel videos to Meta, and all social credentials are encrypted and isolated per user.

What I learned; I learned how OAuth, platform publishing APIs, encrypted token storage, deterministic risk scoring and tamper-evident audit chains work together in a real multi-user product.

Challenges; The biggest challenges were platform verification requirements, Instagram’s need for publicly accessible video URLs, asynchronous Reel processing and implementing secure Bluesky OAuth. These constraints shaped Pevier’s fail-safe design: if anything fails, nothing is published.

whats next; Pevier will add YouTube and X integrations, configurable organization policies, team approval workflows, scheduled publishing and smarter risk detection. I also plan to complete platform reviews so anyone can securely connect their own accounts.