Not a chatbot. Not a script. A fully autonomous AI agent that researches, creates, and publishes content — all by itself.
This project is a fully autonomous AI agent workflow built on Make.com that:
- 🔍 Perceives — Scans the web for the latest Agentic AI news & trends
- 🧠 Reasons — Google Gemini AI thinks and crafts a sharp, on-brand post prompt
- 🎨 Creates — Picsart generates a stunning visual with zero human input
- 📤 Acts — Publishes the post directly to LinkedIn. Autonomously.
No commands. No supervision. No manual work.
This is what Agentic AI actually looks like in the real world.
┌─────────────────────────────────────────────────────┐
│ AGENTIC AI WORKFLOW │
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Web │───▶│ Google │───▶│ Picsart │ │
│ │ Search │ │ Gemini │ │ AI Art │ │
│ │ (Scan) │ │ (Reason) │ │ (Create) │ │
│ └──────────┘ └──────────┘ └────┬─────┘ │
│ │ │
│ ┌───────▼──────┐ │
│ │ LinkedIn │ │
│ │ Auto-Post │ │
│ │ (Publish) │ │
│ └──────────────┘ │
└─────────────────────────────────────────────────────┘
| Tool | Role |
|---|---|
| Make.com | Workflow automation & orchestration engine |
| Google Gemini AI | Content reasoning & prompt generation |
| Picsart AI | Autonomous visual generation |
| LinkedIn API | Direct post publishing |
| Web Search Module | Real-time trend scanning |
The agent uses Make.com's HTTP/Web Search module to autonomously scan for trending Agentic AI content. It pulls headlines, summaries, and engagement signals — no human query needed.
Raw trend data is passed to Google Gemini, which:
- Analyzes what's resonating in the space
- Decides on the most valuable angle
- Crafts a full LinkedIn post + image generation prompt
Gemini's image prompt is sent to Picsart's AI art generation API, which returns a high-quality, on-brand visual — entirely machine-generated.
The post (text + image) is automatically published to LinkedIn via the LinkedIn API module in Make.com. The agent acts without waiting for approval.
agentic-ai-workflow/
├── README.md ← You are here
├── LICENSE
├── docs/
│ ├── SETUP_GUIDE.md ← Full setup instructions
│ ├── WORKFLOW_WALKTHROUGH.md ← Detailed module breakdown
│ └── SCREENSHOTS/ ← Make.com blueprint screenshots
├── assets/
│ └── workflow-diagram.png ← Architecture visual
└── .github/
└── workflows/ ← (Optional) CI/CD configs
- A Make.com account (free tier works for testing)
- Google AI Studio API key → Get one here
- Picsart API key → Get one here
- LinkedIn Developer App with
w_member_socialpermission
See the full Setup Guide → for step-by-step instructions.
- Import the Make.com blueprint (JSON file in
/docs) - Connect your API credentials in Make.com's connection manager
- Configure your LinkedIn profile ID
- Set your preferred posting schedule
- Activate the scenario — the agent runs itself from here
We've moved past "AI that assists."
This is AI that executes.
| Old Model | Agentic AI |
|---|---|
| You prompt → AI responds | AI perceives → AI decides → AI acts |
| Manual trigger | Autonomous trigger |
| Single-step | Multi-step reasoning loop |
| Output: text | Output: real-world action |
The shift from prompt-based AI to Agentic AI is the biggest leap in tech right now. This project is a working demonstration of that shift.
Screen recording of the live workflow running autonomously — from web scan to LinkedIn post.
(See screen recording in project assets)
Ideas, improvements, and forks are welcome. If you build something on top of this:
- Fork the repo
- Create a feature branch (
git checkout -b feature/your-idea) - Commit your changes (
git commit -m 'Add: your feature') - Push and open a Pull Request
This project is licensed under the MIT License — see LICENSE for details.
Syeda Zainab Kamal
"We've moved past AI that assists. This is AI that executes."