Anil-matcha/Open-AI-Design-Agent

Open-source AI design agent — alternative to Lovart AI, Runway Agent, Luma Labs Agent, Krea Agent, Pika Agent, Galileo AI, Magic Patterns. Autonomous multi-step creative for posters, social, brand kits, ad creatives, and video. Free, MIT-licensed, self-hosted.

Open AI Design Agent – Quick Overview

What it is – An open‑source, self‑hostable “design agent” that turns a plain‑language brief into a full set of visual assets (posters, social graphics, brand kits, video ads, etc.). It does this by running an autonomous planning loop that selects the most suitable image‑ or video‑generation model for each asset and stitches the results together.

Why it matters – Commercial tools such as Lovart, Runway Agent, Luma Labs Agent or Krea charge subscriptions and hide the model‑selection logic. This project offers the same agentic workflow for free under an MIT licence, lets you keep data on‑premise, and gives you full visibility into every planning step.

Core Features (as described in the README)

Feature	What you get
Design Agent (Chat mode)	Write a brief (e.g., “launch kit for a sustainable coffee brand”) → the agent creates a deliverable list, picks a model for each item, generates assets in dependency order, and returns a downloadable kit.
Design Studio	One‑click generation of static assets: posters, flyers, logos, brand boards, UI mockups, packaging, etc.
Motion Studio	Generates animated logos, short video ads, social‑media cut‑downs (9:16, 1:1, 16:9) using text‑to‑video or image‑to‑video models.
Multi‑image reference	Up to 14 reference images can be supplied for style transfer, product placement, or character consistency.
Brand‑kit conditioning	Palette, fonts, logo and tone are persisted across every generation in a campaign.
Edit mode	Natural‑language edits (recolor, inpaint, expand, remove elements) using image‑to‑image models.
Templates library	Ready‑made recipes for Instagram posts, LinkedIn banners, YouTube thumbnails, business cards, t‑shirts, etc.
Vectorize & export	Raster outputs can be turned into SVG, backgrounds removed, and upscaled to 4K‑ready files.
Workflow Studio	A visual node‑based editor (powered by the separate Vibe Workflow engine) to build repeatable multi‑step pipelines without writing code.
Model coverage	Orchestrates 200+ image and video models (Flux 2 Pro, Nano Banana 2, Ideogram v3, Recraft v3, Midjourney v7, Kling, Sora, Veo, Runway, Luma Ray2, etc.).
Inspectable agent loop	Every plan, model choice and intermediate output is visible and can be edited or forked.
Self‑hostable	Run locally via Node.js + FastAPI or with Docker Compose; your API key to Muapi.ai is the only external credential needed.

How it works (high‑level flow)

Plan – The agent parses the natural‑language brief and creates a structured list of assets to produce.
Route – For each asset it selects the most appropriate model (e.g., Recraft v3 for logos, Ideogram v3 for typographic posters, Kling for video).
Execute – Calls the Muapi.ai gateway to run the chosen model, passing along brand‑kit data and any reference images.
Assemble – Packages all outputs into a downloadable folder, preserving the original order and metadata.

You can pause at any step, swap models, lock a result, or edit an intermediate image before continuing.

Getting started

Step	Command / Action
Clone	`git clone https://github.com/Anil-matcha/Open-AI-Design-Agent.git`
Install client	`npm run install:all`
Run client	`npm run dev:app` (http://localhost:3000)
Install server	`cd server && pip install -r requirements.txt`
Run server	`uvicorn app.main:app --reload --port 8000`
Docker	`docker compose up --build` (starts both client and server)
First launch	Enter your free Muapi.ai API key in the UI.

A hosted, no‑install version is also available at https://muapi.ai/assistant.

Who might use it?

Design freelancers who want a free, scriptable alternative to paid AI design suites.
Start‑ups / agencies needing to keep brand assets on‑premise for privacy.
Developers who want to extend the agent with custom planners, models or UI components.
Researchers exploring multi‑modal generation pipelines and agentic workflows.

License & Community

License: MIT – free for commercial and personal use.
Support: Reddit community, Discord server, and a collection of related open‑source projects (Open‑Generative‑AI, Vibe Workflow, AI‑Voice‑Agent, etc.).

Bottom line: Open AI Design Agent is a full‑stack, open‑source platform that automates the end‑to‑end creation of static and motion marketing assets by orchestrating a large catalog of generative models, while keeping everything transparent and self‑hostable.