AutoPost AI: An Autonomous Social Media Operations Pipeline Using Multi-Agent LLMs
TL;DR
AutoPost AI is an open-source multi-agent system that autonomously handles a full social media content workflow — from researching trends to writing posts, sourcing images, scheduling content, and emailing a weekly report. Built with LangGraph, LangChain, and Python, powered entirely by free-tier APIs.
Overview
You run one command. A pipeline of 6 specialized AI agents handles the rest — research, planning, writing, image sourcing, scheduling, and reporting — with zero human involvement after setup. The system is built as a LangGraph StateGraph where each agent passes typed, validated data to the next through shared state.
Target Audience
This project is for AI/ML students learning multi-agent system design, developers looking for a practical LangGraph reference implementation, and anyone who wants to automate social media content at zero cost. Basic Python knowledge is enough to get started.
Why It Matters
AutoPost AI is fully autonomous end-to-end, completely free to run using only free-tier APIs, and uses a real multi-agent architecture with 6 agents, typed Pydantic inter-agent communication, and explicit state management via LangGraph.
Architecture
The pipeline executes in this order: load history, Trend Researcher, Content Planner, Writer, Image Sourcer, Scheduler, Reporter, then sends the weekly email and ends. Each agent has a distinct role and communicates with the next through Pydantic schemas defined in models/schemas.py, ensuring validated data at every step.
Setup
Requirements: Python 3.9+, Gmail with 2FA, and free accounts on Groq, Serper, and Unsplash.
Clone the repo, create a virtual environment, install dependencies with pip install -r requirements.txt, then create a .env file with your API keys: GROQ_API_KEY, SERPER_API_KEY, UNSPLASH_ACCESS_KEY, GMAIL_SENDER, GMAIL_APP_PASSWORD, and GMAIL_RECIPIENT. Finally edit config.py to set your brand details.
Usage
Run python main.py and type "start" when prompted. The pipeline takes 3 to 5 minutes to complete and outputs a scheduled content calendar saved to memory/content_store.json and a weekly report delivered to your inbox.
Limitations
Posts are not published directly — they are saved locally for manual publishing. Image selection is text-based only. The system is subject to Groq's 100k daily token limit on the free tier.
Tech Stack
LangChain, LangGraph, Groq API (Llama 3.3 70B), Pydantic, Tenacity, Serper API, Unsplash API, Gmail SMTP.