Dzhero

• 

  Agent Studio turns one real signal into grounded evidence and a manager-reviewed content package.

• 

  Independent quality gate checks grounding, brand fit, feasibility, hook strength, and mechanic fidelity.

• 

  Creative Producer generates a full production script, alternative directions, and shot-by-shot Reel timing.

• 

  One approved concept expands into a seven-day content plan ready for the calendar.

• 

  Transparent agent activity shows each specialist step from evidence analysis to final approval.

Inspiration

Short-form content moves fast, but most small brands do not have a full creative team to analyze trends, adapt them safely, write scripts, and build a weekly content plan.

Dzhero was inspired by that gap. A business owner can see a strong Reel, TikTok, or Short and understand that it works, but turning that signal into something original, grounded, and useful for their own brand is still hard.

What it does

Dzhero is an AI Agent Studio for short-form content planning.

It starts with one real video signal and turns it into a full creative package:

• grounded evidence from the source video
• the transferable content mechanic
• brand adaptation
• several creative directions
• an independent quality review
• a production-ready Reel script
• a seven-day content plan

The goal is not to copy trends. The goal is to understand why a signal works and help a brand create its own version.

How we built it

Dzhero uses a manager-led agent workflow. Jeryk, the manager agent, coordinates a team of specialist agents:

• Trend Analyst
• Video Evidence Analyst
• Brand Strategist
• Creative Producer
• Critic
• Content Planner

The agents use structured outputs, evidence references, quality gates, and a transparent activity log so the user can see how the final result was produced.

We used Codex throughout the build process to design the agent workflow, iterate on prompts, implement the app, fix bugs, prepare the README, and polish the final demo flow.

GPT-5.6 was used inside Codex as the main development partner for planning, implementation, debugging, and submission preparation.

The project is built with React, Node.js, Express, PostgreSQL, Gemini video evidence, OpenAI Agents SDK, and GPT-5.6-assisted development.

Challenges

The hardest part was making the system feel useful instead of generic.

We had to make sure the agents stayed grounded in the original source, did not invent unsupported claims, respected the brand context, and produced something practical enough to shoot.

Another challenge was orchestration: each specialist agent has to pass useful context to the next step without exposing private prompts, provider payloads, or hidden reasoning.

What we learned

We learned that agent workflows become much more valuable when each agent has a clear job, a strict output contract, and a visible reason to exist.

The best results came from combining creative generation with verification: evidence first, brand adaptation second, creative production third, and quality review before the content plan.

What's next

Next, Dzhero can expand into team approvals, performance feedback, deeper brand memory, and more reliable discovery of fresh market signals.

The long-term vision is a workspace where small teams can go from real market signals to planned, shootable content without starting from a blank page.

Built With

• ai-agents
• content-planning
• express.js
• gemini
• generative-ai
• gpt-5.6
• node.js
• openai-agents-sdk
• postgresql
• railway
• react
• short-form-video
• vite
• zod