Show HN: I built a social media management tool in 3 weeks with Claude and Codex

TL;DR Highlight

**SoloDev built a Buffer/Sendible alternative open-source social media management platform in 3 weeks by leveraging AI coding tools like Claude Opus and OpenAI Codex.**

Who Should Read

Developers looking to apply AI coding tools to real-world production projects, or those seeking a self-hosted social media management tool instead of SaaS options like Buffer or Sendible.

Core Mechanics

This project resulted in a full-stack SaaS-level platform completed solo in 3 weeks, featuring integrations with 12 social platforms (Facebook, Instagram, LinkedIn, TikTok, YouTube, Pinterest, Threads, Bluesky, Google Business Profile, Mastodon, etc.), multi-tenant authentication, encrypted credential storage, background job processing, approval workflows, and a unified inbox.
Detailed spec documents, architecture documents, and style guides were created *before* writing any code and were all made public. The developer emphasized that 'this pre-planning phase was everything; without it, the AI agents would be a mess.'
Specs were divided into tasks executable in parallel and tasks requiring sequential processing to form multi-agent workflows. This categorization was key to the AI coding workflow.
Claude Opus 4.6 (Claude Code) was used for maintaining large contexts and making architectural decisions across multiple files, particularly excelling at cascading changes across models, views, and templates when restructuring the permission system.
OpenAI Codex 5.3 was leveraged to review implemented code, identify security issues, and fix bugs. Token consumption between the two models was roughly equal.
AI excelled in areas like standard CRUD operations (Django models/views/serializers), provider modules for well-documented APIs (Facebook, LinkedIn), Tailwind layouts and HTMX interactions, test generation, and large-scale refactoring across files.
AI faltered in areas with poor documentation or unique upload flows, such as TikTok’s Content Posting API, repeatedly generating incorrect code with confidence. A multi-tenant permission logic bug exposed data across workspaces, even passing tests, making it more dangerous.
OAuth edge cases (token refresh, permission revocation, platform-specific error codes) and background task orchestration (retry logic, rate limit backoff, error handling) all required manual implementation. AI handles 'happy paths' well but is vulnerable in defensive coding.

Evidence

"Some responses questioned the credibility of the '3 weeks of vibe coding' claim, with one commenter noting that similar open-source tools had poor quality and self-hosting experiences, leading them to prefer paid premium products. The claim signaled a lack of battle testing and potential maintenance issues."

How to Apply

When building production-level projects with AI agents, create detailed spec documents and architecture documents *before* writing code, and separate tasks into those executable in parallel and those requiring sequential processing. Benchmark your spec writing against the specs published in the brightbean-studio repository’s development_specs folder.
When configuring multi-agent workflows, use Claude Opus (or a high-performance model) for initial architecture design and large-scale refactoring, and a separate model (Codex, etc.) for reviewing implemented code and identifying security issues. This can catch more bugs and security vulnerabilities than using a single model.
When implementing multi-tenant environments, do not test AI-generated permission logic only in a single-tenant environment. AI can create data cross-tenant exposure bugs that still pass unit tests, so write separate integration tests explicitly including multi-tenant scenarios.
Agencies or small teams currently paying $100-$300/month for SaaS solutions like Buffer, Sendible, or SocialPilot can deploy brightbean-studio to their own VPS with Docker Compose or use Render/Railway one-click deployment for the same functionality at no cost. However, be sure to review the AGPL-3.0 license conditions (requirement to open-source modifications).

Terminology

멀티 테넌트(multi-tenant)An architecture where a single app instance simultaneously manages data for multiple independent customers (tenants), ensuring data isolation.

HTMXA lightweight library that enables AJAX requests and dynamic updates using only HTML attributes, minimizing the need for JavaScript. Allows for dynamic UIs without React.

AGPL-3.0An open-source license that requires any modifications to the software to also be open-sourced if the software is operated as a network service.

레이트 리밋 백오프(rate-limit backoff)A retry logic used when an API call limit is exceeded, involving waiting a period of time before retrying, with the wait time increasing progressively to reduce server load.

백그라운드 잡(background job)A task processed asynchronously in a separate process, not immediately in response to a user request. Used for tasks that need to be executed later, such as scheduling social media posts.

퍼스트파티 API(first-party API)An API directly and officially provided by the platform (Facebook, Instagram, etc.), maintaining data sovereignty without relying on third-party aggregators.