How to vibe codeOutstand

Prerequisites

• Node.jsfree

  Runtime for the TypeScript backend service and MCP server

• GitHubfree

  Repository hosting and deployment triggers

AI coding tools

• Claude Code

  Best-in-class multi-file scaffolding and handling complex backend API glue logic

• Cursor

  Ideal for inspecting API request/response diffs and editing TypeScript handlers

Build guide

1. 01Project Scaffolding and Database Schema

   Initialize a TypeScript backend with Hono, Turso for SQLite storage, and better-auth for API key generation.

   Scaffold a new TypeScript backend project using Hono and Drizzle ORM configured for Turso. Create database tables for organizations, api_keys, social_accounts (storing encrypted tokens, network name, account metadata), posts (storing content, scheduled_at, published_at, idempotency_key), and post_containers (for multi-image/thread payloads). Implement authentication middleware that reads the Authorization: Bearer <API_KEY> header, validates it against the api_keys table, and injects the organization ID into the request context. Ensure proper error handling returning standard 401 JSON payloads for missing or invalid keys.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

2. 02OAuth Credential Vault and Social Account Connection

   Implement OAuth redirect endpoints and credential vault storage for connecting social networks.

   Build social account connection endpoints in Hono. Create POST /v1/social-accounts/{network}/{orgId} that initiates OAuth redirects for X, LinkedIn, and Meta platforms, and a POST endpoint for Bluesky app passwords. Store returned OAuth access tokens, refresh tokens, and expiration timestamps securely encrypted in the database. Implement GET /v1/social-accounts to list connected accounts filtered by organization ID, and DELETE /v1/social-accounts/{id} to disconnect accounts and revoke local tokens.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

3. 03Presigned URL Media Processing Pipeline

   Implement the two-step presigned upload workflow targeting Cloudflare R2 storage.

   Implement the media upload pipeline using S3-compatible presigned URLs pointing to Cloudflare R2. Create POST /v1/media/upload that accepts filename and content_type, generates a unique media ID, and returns a PUT presigned upload URL with an expiration window. Create POST /v1/media/{id}/confirm to verify file size and activate the media record, returning a public CDN URL. Store media records with a status enum (pending, active, expired) and ensure cleanup workers remove unconfirmed items after 60 days.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

4. 04Unified Post Creation and Idempotency Engine

   Build the unified post creation endpoint with multi-platform fan-out and idempotency tracking.

   Build POST /v1/posts/ supporting unified post creation across multiple target accounts. Parse incoming JSON bodies containing containers (content and media arrays) and account specifiers (IDs, network names, or usernames). Implement idempotency checking via the Idempotency-Key request header: cache responses for 24 hours, return cached results with Idempotency-Replayed: true on retry, return 409 if a request is currently in flight, and return 422 if the key is reused with a differing body. For immediate posts, trigger asynchronous background dispatch tasks to upstream platform APIs.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

5. 05Scheduler and Background Worker Pipeline

   Implement a polling worker that processes scheduled posts when their timestamp arrives.

   Implement a background cron worker that runs every minute to query the posts table for records where scheduled_at <= NOW() and published_at IS NULL. For each due post, fan out execution to the respective platform adapter handlers (X API, LinkedIn API, etc.), handle per-platform rate limits with exponential backoff and retries, capture native platform post IDs or error messages into the per-account result array, and update published_at once processing concludes.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

6. 06Model Context Protocol (MCP) Server Integration

   Expose an authenticated Model Context Protocol server enabling AI agents to interact with social endpoints.

   Implement a remote Model Context Protocol (MCP) server endpoint using the Model Context Protocol SDK in TypeScript. Expose executable tools including create_post, list_posts, get_post_analytics, list_social_accounts, and upload_media. Ensure the MCP server validates API keys provided in the initialization handshake, translates LLM natural language parameters into standard internal API service calls, and returns clean, structured markdown tool outputs for AI coding assistants like Claude and Cursor.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

Cost vs paying for Outstand